Identification of Organisms: Algorithm and Tools

Error in Identifying a Pest → Wrong Pesticide

Imagine the situation: an agronomist finds on cereal crops an insect that outwardly resembles cereal aphid. A decision is made to apply an insecticide against aphids. However, on closer analysis it turns out that this is not an aphid but a thrips larva or a leafhopper — an organism with different biology, different susceptibility to active ingredients, and different treatment timing. The result is crop loss, unjustified pesticide costs, and, especially important, unnecessary pesticide load on the agrocoenosis.

Barhate et al. (2024), in their systematic review of plant identification methods, emphasize that “manual analysis of these parameters at the same time is a significantly labor-intensive task, and sometimes even experts may find it difficult to assess them correctly.” This problem is aggravated in field conditions, where access to specialized literature and equipment is limited.

Error in Identifying a Pathogen → Crop Loss

In phytopathology the situation is even more critical. Many disease agents have similar symptoms in the early stages of infection. For example, as noted by Garrido et al. (2016) in the chapter on fungal diseases of strawberry, “many strawberry disease agents produce similar symptoms, and it is important to be able to distinguish different species.” The authors give a concrete example: leaf spot caused by Diplocarpon earlianum is often confused with symptoms of spot caused by Mycosphaerella fragariae, and symptoms of the latter are mistakenly identified as Phomopsis leaf blight caused by Phomopsis obscurans (Garrido et al., 2016).

The consequences of such an error are diverse:

• Wrong choice of fungicide (different pathogens may require different active ingredients)
• Wrong treatment timing (life cycles differ)
• Inability to predict an epidemic (different pathogens have different development dynamics)

In the review by Tso et al. (2021) on diagnosis of clubroot of crucifers (Plasmodiophora brassicae), it is emphasized that “early, preventive determination of the pathotype is necessary to maximize productivity and promote sustainable crop production.” The authors note that the spread of resistant cultivars has led to the emergence of new, virulent pathotypes, and only accurate identification allows correct selection of a cultivar and protection strategy.

Quarantine Incidents

A special category consists of quarantine organisms. Here an error in identification may have not only agronomic but also legal consequences. As noted in the review by Cubero et al. (2024), “accurate detection and identification of a pathogen are of great importance because they can provide crucial information about its ecology, potential threat to crops, and information about the introduction of appropriate management methods.”

Imagine the situation: during inspection of an imported batch of seeds, a fungal pathogen is detected. If it is mistakenly identified as a non-quarantine species, the batch will be admitted into the country, which may lead to the introduction of a dangerous agent. Conversely, overdiagnosis (false determination of quarantine status in a harmless organism) leads to unjustified economic losses for the importer and trading partners.

In the diagnostic manual by Tso et al. (2021), it is directly stated that “early determination of the pathotype allows farmers growing canola to choose cultivars with appropriate resistance and ensure the best protection against the disease.” This is especially important under conditions when, as the authors note, “repeated cultivation of resistant cultivars has created selective pressure that has led to virulence shifts in the pathogen, promoting the spread of new virulent pathotypes.”

Levels of Taxonomic Detail

In the agrosphere, identification tasks may require determination at different levels:

Family or genus level — often sufficient for preliminary assessment. For example, upon detecting an insect on crops, it is enough to determine that it is a representative of the family Aphididae (aphids) in order to decide whether more detailed analysis is necessary. As noted in the manual on integrated plant protection (Albajes and Madeira, 2019), “when a technician observes insects or mites on a crop, the first question that arises is: ‘what is this?’ The exact answer to this question is the key starting point for finding an effective solution if the population grows to the level of a pest.”

Species level — necessary for most practical decisions. It is at this level that economic thresholds of harmfulness are determined, pesticides are selected, and protection strategies are developed. As emphasized by Tso et al. (2021), “accurate identification to species allows farmers to choose cultivars with appropriate resistance and ensure the best protection against the disease.”

Level of intraspecific categories — races, pathotypes, biotypes. In some cases even identification to species is insufficient. For example, for Plasmodiophora brassicae (the causal agent of clubroot of crucifers), many pathotypes are distinguished, which are differentiated by virulence on a set of differential plants (Tso et al., 2021). To choose a resistant cultivar, it is necessary to know which pathotype is present in the soil.

Strain or isolate level — in epidemiological studies and in the development of biological preparations. For example, for Bacillus thuringiensis, different strains possess different insecticidal activity against different groups of insects (Albajes and Madeira, 2019).

Factors Determining the Required Level

The choice of level of detail is determined by several factors:

1. Purpose of the study or practical task. If the goal is general phytosanitary monitoring, identification to family may be sufficient. If the goal is choosing a pesticide, identification to species is necessary. If the goal is breeding for resistance, identification to pathotype may be required.

2. Available resources. Identification to species may require consultation with a specialist systematist, access to reference collections, or a molecular laboratory. If such resources are unavailable, one has to limit oneself to a lower level of detail.

3. Economic consequences of error. The higher the cost of error, the higher the required level of detail. For quarantine organisms, even the probability of error must be minimized, which requires the use of the most reliable methods.

4. Nature of the object. Some groups of organisms are objectively more difficult to identify. As noted in the review on DNA barcoding of plants (Letsiou et al., 2023), “in plants the use of a universal DNA barcode such as COI, which is used in animals, has still not been achieved.” Different groups require different markers and different approaches.

Formulation of the Task: Practical Recommendations

When setting an identification task, it is recommended to answer the following questions:

• What is known about the sample? (origin, host, symptoms, conditions)
• What level of identification is necessary? (family, genus, species, race)
• What methods are available? (morphology, microscopy, molecular methods)
• What limitations exist? (time, budget, access to equipment)
• What are the consequences of error? (economic, legal, environmental)

A clear formulation of the task allows avoiding two extremes: insufficient detail (when the result does not allow a decision) and excessive detail (when the costs of identification are not justified by practical necessity).

General Principles of Collection

In the manual on preparing herbarium specimens (Judd et al., 2018, Appendix 2), the main principles of plant collection are formulated:

1. Completeness. “Collected plants should represent the variation observed in the population. Several specimens may be needed to adequately document the observed pattern of variation” (Judd et al., 2018).

2. Representativeness. “Choose well-developed, non-diseased plants. If possible, the whole plant should be collected, including underground parts; never collect only one flower or leaf!” (Judd et al., 2018).

3. Presence of diagnostic characters. “Do not collect specimens that lack flowers and fruits; such material will be difficult to identify” (Judd et al., 2018).

4. Documentation. Each specimen must be supplied with a label containing complete information: place of collection, date, collector, characteristics not preserved upon drying (color, odor, dimensions).

Specificity of Collecting Different Groups of Organisms

Plants. For herbarium preparation, plants are collected whole or in parts, which are then dried in a press. As noted by Judd et al. (2018), “the plant should be placed in folded newspaper and arranged so that the upper and lower surfaces of leaves, flowers, and fruits are visible.” For succulent plants and large fruits, preliminary fixation in alcohol or hot water may be required.

Insects. For insects, dry mounting on entomological pins is usually used, or preservation in alcohol (70–96% ethanol). The choice of method depends on the group: Coleoptera and Lepidoptera are usually mounted on pins, while soft-bodied forms (larvae, aphids, mites) are preserved in alcohol.

Fungi and microorganisms. For phytopathogenic fungi, both drying (herbarization of affected organs) and preservation of pure cultures on nutrient media are used. As noted by Garrido et al. (2016), “the development of molecular methods has led to the generation of a large amount of information about these diseases,” which requires preservation of material suitable for DNA analysis.

Soil samples. For analysis of soil organisms (nematodes, protozoa, microarthropods), special sampling methods are used. Lentendu et al. (2022) describe the procedure: “Using a sampler, take a total of 9 samples (about 250 g each), evenly distributed along the outer edge of a circular plot or evenly scattered within a square plot with a radius/side length of 1 to 10 m.”

Fixation and Preservation

The choice of fixation method is determined by the goals of subsequent analysis:

For morphological analysis — drying (herbarization) or preservation in alcohol. It is important to ensure preservation of diagnostic characters.

For molecular analysis — rapid fixation in ethanol (96–100%) or silica gel. As noted in the manual on seed collection (Hopkins, 2015), “silica gel (for collecting leaves for DNA sequencing)” is a standard element of field equipment.

For microbiological analysis — preservation under sterile conditions, often using special transport media.

It is important to understand that fixation methods optimal for one type of analysis may be unsuitable for another. For example, as noted by Judd et al. (2018), samples treated with alcohol for temporary preservation “will usually be unsuitable for subsequent chemical (including DNA) analysis.”

Ethical and Legal Aspects of Collection

Collection of organisms is regulated by legislation. Judd et al. (2018) emphasize: “Many plants that were once common are now rare, primarily because of habitat destruction, overexploitation by humans, or competition with exotics. Collectors must be conscious of conservation issues. Rare or unusual plants should usually not be collected; photographing a plant is a good alternative in such circumstances.”

In addition, “the laws of many countries protect native plant species, and they should be collected only with appropriate permits. Permits are required for collection in city, county, state, and national parks and forests” (Judd et al., 2018). For collection on private lands, permission from the landowner is required. International trade in plants is regulated by the Convention on International Trade in Endangered Species of Wild Fauna and Flora.

Macroscopic Analysis

At this stage, characters visible to the naked eye or with a hand lens are studied:

For plants — Sharma (2009) lists the following characteristics that should be noted before identification:

• “Determine the herbaceous or woody, annual or perennial nature of the specimen”
• “Note the leaf type, phyllotaxis, and venation”
• “Note the presence or absence and type of stipules on young shoots”
• “Is there milky or colored sap in the leaf, stem, or other parts of the plant”
• “Note the distribution and types of surface coverings (i.e., hairs, trichomes, spines, etc.)”

For insects — morphological characters include the structure of the mouthparts, wing venation, structure of antennae, legs, and genitalia. As noted in the IPM manual (Albajes and Madeira, 2019), “identification of insects and mites is usually carried out using taxonomic keys. Keys are arrangements of related taxa put in clusters. Morphological characters — supplemented sometimes by anatomy, appearance, or behavioral features — are mainly used to separate individuals into clusters.”

For fungi — macroscopic characters include the type of lesion, color, texture, and presence of sporulation. Garrido et al. (2016) describe symptoms of various fungal diseases of strawberry, emphasizing that “some pathogens cause very characteristic leaf symptoms; for example, Sphaerotheca macularis forms white patches of mycelium on the abaxial surface of the leaf. Other species cause similar, even identical symptoms.”

Microscopic Analysis

For many groups of organisms, macroscopic characters are insufficient, and microscopy is required.

For fungi — microscopy allows revealing the shape and size of spores, the structure of conidiophores, and the presence and type of septa in the mycelium. Garrido et al. (2016) describe various staining methods: “Resting spores can be distinguished from root tissue using the triple staining method; resting spores stain blue, whereas host tissue stains pink or purple.” Other methods include staining with lactophenol cotton blue to detect chitin, methylene blue to visualize spores inside plasmodia, and Harris hematoxylin to observe primary plasmodia and zoospores.

For insects — microscopy is necessary for studying genitalia (often the only reliable character for distinguishing closely related species), as well as for studying small structures (setae, sensilla).

For nematodes — microscopy is the main method of identification, since these organisms have dimensions requiring magnification. Lentendu et al. (2022) note that “deep taxonomic understanding of nematode communities requires deep expertise and time and therefore is expensive.”

Anatomical Methods

For some groups, study of internal structure is required:

For plants — anatomical sections can reveal features of stem, leaf, and root structure that have diagnostic significance. Sharma (2009) recommends “make a cross section of the ovary, count the number of locules, the number of ovules in the locule, and also note placentation.”

For wood — anatomical characters (arrangement of vessels, presence and type of rays) are used to identify woody species.

For insects — anatomical methods are used less often, but can be useful for studying internal structures.

Biochemical and Physiological Tests

In some cases, biochemical characters are used for identification:

For bacteria — tests for enzymatic activity, ability to assimilate various substrates, serological reactions.

For fungi — tests for ability to assimilate various carbon and nitrogen sources, enzyme production.

For plants — chemotaxonomic characters (composition of flavonoids, alkaloids, essential oils).

As noted in the review on pathogen diagnostics (Cubero et al., 2024), “serological methods have not always been replaced by nucleic acid-based technologies, especially when antisera, antibodies, and specific tests are available, sufficiently accurate, and sometimes already commercialized.”

Dichotomous Keys

Principle. A dichotomous key is a sequence of pairs of mutually exclusive statements (couplets). Each pair is called a couplet, and each statement is called a lead. The user successively chooses one of two variants until reaching the name of a taxon.

Sharma (2009) defines a key as “an artificial device or analytical means by which a choice is provided between two contradictory statements, leading to the acceptance of one and rejection of the other.”

Types of dichotomous keys:

1. Yoked (indented) keys — each pair of couplets is indented a fixed distance from the left margin. Example from Sharma (2009):

1.1. Fruit a group of achenes; flowers without spurs

1.1.1. Petals absent:

• Sepals usually 4; involucre absent → Clematis
• Sepals usually 5; involucre present → Anemone

1.1.2. Petals present → Ranunculus

2. Bracketed keys — couplets are arranged sequentially, each leading to the next number. Example:

• (1) Fruit a group of achenes; flowers without spurs → (2)
• (1) Fruit a group of follicles; flowers with spurs → (4)
• (2) Petals absent → (3)
• (2) Petals present → Ranunculus

Advantages of dichotomous keys:

• Simplicity of use
• Clear logical structure
• Wide distribution

Limitations:

• If a character used in the key is absent or unclear in the specimen, identification becomes difficult
• The key may be unsuitable for incomplete material
• Dichotomous keys often require knowledge of specific characters that may be unavailable

As noted by Sharma (2009), “the main disadvantage is that if a particular character is unclear in the specimen for identification, the only available option is to make an assumption about that particular feature.”

Polythetic (Multi-Access) Keys

Principle. A polythetic key (also called a multi-access key, polyclave, synoptic key) allows the user to choose characters in any order from a set of available ones.

Pandey (2021) defines a polythetic key as “a list of character states from which the user selects all states corresponding to the specimen and then passes the list to a computer program that matches the list of characters with a database of taxa and characters and returns the identification to the researcher based on matches.”

Types of polythetic keys:

1. Edge-punched cards — each card corresponds to a taxon, and holes along the edges correspond to characters. If a character is present, the hole is notched. For identification, cards are stacked and a needle is passed through the hole corresponding to the selected character; cards with a notched hole fall out.

2. Body-punched cards — each card corresponds to a character, and holes correspond to taxa. Cards are selected according to characters present in the specimen, stacked, and viewed against light; the matching hole indicates the taxon.

3. Computer polythetic keys — implemented as software (for example, DELTA, Lucid).

Advantages of polythetic keys:

• Flexibility: any available characters can be used
• Possibility of identifying incomplete material
• Possibility of obtaining a list of possible taxa when information is insufficient

Limitations:

• Require more time for preparation
• Computer implementations require access to equipment
• May be less intuitive for beginners

As noted by Pandey (2021), “polythetic keys are convenient, inexpensive, and compact, but few are available. Preparing these cards and selecting characters take time, but these keys can be used even if the specimen lacks some traditional key characters.”

Computerized and Online Identification Tools

Modern technologies offer new possibilities for identification:

Computer keys — programs implementing dichotomous or polythetic keys. Sharma (2009) notes that “a computer can be programmed for identification only if the initial algorithm was successfully prepared by an expert systematist.”

Online resources — websites providing access to keys, images, and databases. Watson et al. (2015) describe the development of such resources: “The World Wide Web provides a wealth of information about plants, access to floras and checklists, plant images, detailed treatments of particular plant groups, and DNA sequences.”

Mobile applications — smartphone applications using machine learning for automatic identification. Jones (2025) evaluates their effectiveness: “All tested applications improved over the 3 years between 2020 and 2023 by approximately 20 percentage points in terms of the percentage of samples correctly identified to species or family.”

However, as we will see in Lecture 6, automatic identification tools have serious limitations and cannot replace the critical thinking of a specialist.

Comparison with Reference Specimens

Herbarium collections. The most reliable method of verification is comparison with type specimens or verified specimens in a herbarium. Judd et al. (2018) emphasize: “Herbarium specimens form the basis of much of our understanding of patterns of variation in nature. They document the morphological variability of populations, species, and higher taxa, their geographic distribution, and ecological characteristics.”

Type specimens. Of particular value are type specimens — specimens to which “the name of a taxon is permanently attached” (Sharma, 2009). Comparison with the type allows ensuring that identification corresponds to the nomenclatural standard.

Digital collections. Modern technologies allow access to digitized images of type specimens via the internet. Sharma (2009) describes virtual herbaria: “A virtual herbarium, also known as an online digital herbarium, is a repository containing digital images and information about preserved plant specimens.”

Consultation with a Specialist

When consultation is necessary. Consultation with a specialist systematist is necessary in the following cases:

• Identification is ambiguous
• The specimen belongs to a complex taxonomic group
• Identification to intraspecific level is required (race, pathotype)
• The result is critically important (quarantine, forensic examination)

How to organize consultation. Pandey (2021) notes that “problem collections can be sent to an expert on the systematics of a particular genus or family or on the flora of a particular region.” It is important to provide the specialist with complete information about the specimen, including photographs, description, and place of collection.

Cooperation with diagnostic laboratories. In the agrosphere there are specialized diagnostic laboratories that carry out identification of pathogens, pests, and weeds. Cubero et al. (2024) emphasize that “accurate detection and identification of a pathogen are of great importance,” and in many countries there are official laboratories accredited for such studies.

Use of Several Sources

To increase the reliability of identification, it is recommended to use several independent sources:

• Several keys (if available)
• Comparison with several reference specimens
• Consultation with several specialists (if possible)
• Combination of morphological and molecular methods

As noted by Sharma (2009), “after using the key it is important to read the corresponding description of the taxon; if the specimen was correctly identified, the description of the identified species should correspond to the plant in hand.”

When Morphology Is Insufficient

Cryptic species. There are groups of organisms whose species are morphologically indistinguishable but genetically isolated. Letsiou et al. (2023) note that “cryptic species are two or more species that are classified as one because they are superficially morphologically similar and therefore cannot be identified using classical taxonomy.”

Incomplete material. When available material does not contain diagnostic characters (for example, insect larvae, sterile plant specimens), molecular methods may be the only way of identification.

Mixed samples. When analyzing mixed samples (for example, soil, plant material with several pathogens), metabarcoding allows identifying all components of the community.

High accuracy required. For quarantine purposes, forensic examination, and breeding programs, maximum reliability may be required, which molecular methods provide.

DNA Barcoding

Main page: DNA barcoding and digital taxonomy

Principle. DNA barcoding is based on the use of standardized short DNA sequences for species identification. Letsiou et al. (2023) define a DNA barcode as “standardized DNA sequences, ideally unique, coding or non-coding, either from the genome of the organism or from its organelles, 400 to 800 base pairs in length, which are used for identification/classification of a group of organisms.”

Markers for different groups:

• Animals: COI (cytochrome c oxidase subunit I)
• Plants: combination of rbcL + matK (CBOL Plant Working Group, 2009)
• Fungi: ITS (internal transcribed spacer)

Letsiou et al. (2023) emphasize that “in plants the use of a universal DNA barcode such as COI, which is used in animals, has still not been achieved,” and that “there is no single locus that would meet CBOL standards, so the use of a combination of barcodes has been proposed.”

Procedure:

1. DNA extraction from the sample

2. Amplification of the target region by PCR

3. Sequencing of the amplicon

4. Comparison with a reference database

Pandey (2021) describes this process: “DNA barcoding includes extraction of genomic DNA from tissue samples collected from an individual organism. This is the first step, followed by amplification of the desired region using polymerase chain reaction. The amplified product is then subjected to sequencing. Bioinformatics tools are used to analyze the obtained sequences.”

qPCR for Pathogen Diagnostics

Principle. Quantitative real-time PCR (qPCR) allows not only detecting but also quantifying the content of a pathogen in a sample.

Application. As noted by Cubero et al. (2024), “in recent years conventional PCR has been replaced by real-time qPCR, which usually has higher sensitivity and relies on automated systems.” For many pathogens, specific primers and probes have been developed, allowing diagnostics with high sensitivity and specificity.

Advantages:

• High sensitivity (detection of single copies)
• Quantitative assessment
• Rapidity
• Possibility of multiplexing (simultaneous detection of several pathogens)

Metabarcoding

Principle. Metabarcoding is a method of analyzing communities of organisms without isolating individual specimens. Lentendu et al. (2022) define it as “a method of identifying multiple species in a single environmental sample using universal PCR primers to simultaneously amplify many DNA barcodes.”

Application:

• Analysis of soil communities (protozoa, nematodes, microarthropods)
• Analysis of the plant microbiome
• Biodiversity monitoring

Advantages:

• Possibility of analyzing complex communities
• Detection of unculturable organisms
• High throughput

Limitations:

• Relative rather than absolute quantification
• Dependence on reference databases
• Possibility of contamination
• Complexity of bioinformatic analysis

Lentendu et al. (2022) emphasize that “metabarcoding cannot provide absolute quantities, only estimates of proportions of pathotypes,” and that “sample contamination is of great concern because of the high sensitivity of NGS platforms.”

Limitations of Molecular Methods

Despite all advantages, molecular methods have limitations:

Contamination. As noted by Cubero et al. (2024), “molecular methods have a limitation: they detect microorganisms in any physiological state or simply inert traces of DNA or RNA molecules from dead microorganisms.” This means that detection of pathogen DNA does not necessarily mean the presence of a viable pathogen.

Low titer. When the pathogen content in a sample is low, molecular methods may give false-negative results.

Mixed samples. When analyzing mixed samples, interpretation of results may be difficult.

Lack of reference sequences. Letsiou et al. (2023) note that “DNA barcodes cannot be a useful identification tool without a comprehensive and reliable reference database.”

Different taxonomic concepts. As noted by Hyam (2015), “scientific, type-based names are not adequate identifiers of taxa.” Different researchers may use different taxonomic concepts, which leads to ambiguity in interpreting results.

Voucher Specimen

Definition. A voucher specimen is a specimen that documents identification and is preserved for subsequent verification.

Requirements for a voucher specimen:

• Completeness (presence of diagnostic characters)
• Correct fixation and preservation
• Complete label

Sharma (2009) describes the information that must be on a herbarium label:

• “Scientific name of the plant: correct and accepted binomial name, including authorship”
• “Place of collection: the place from which the specimen was collected”
• “Altitude above sea level”
• “Latitude and longitude”
• “Date of collection”
• “Collector: name of the person who collected the plant”
• “Voucher/collection number”
• “Field notes: additional information about the plant that may disappear after drying”

Storage. Voucher specimens must be stored under appropriate conditions. Sharma (2009) notes that “herbarium specimens will be preserved indefinitely with careful handling. However, they are susceptible to damage by insects, fungi, and fire, and should be stored in specialized cabinets in a climate-controlled environment.”

Photographs

Significance. Photographs are an important addition to a voucher specimen, especially for documenting characters that are not preserved upon drying (color, shape, arrangement).

Requirements. Watson et al. (2015) emphasize the importance of standardization: “For some organisms, video is becoming a useful new method of collection... these data can be archived and serve as vouchers when a whole individual is needed for DNA extraction.”

Recommendations:

• Photograph the plant as a whole and individual organs
• Use a scale ruler
• Record color (use a color scale)
• Record GPS coordinates
• Take several images from different angles

Metadata

Definition. Metadata is information about a specimen that is not its physical part but is necessary for interpretation.

Composition of metadata:

• Taxonomic identification (name, author, identifier)
• Date and place of collection
• Ecological data (habitat, host, substrate)
• Method of collection and fixation
• Relation to other specimens (duplicates, cultures)
• Information about molecular data (accession numbers in GenBank)

Standards. Hyam (2015) emphasizes the importance of using globally unique identifiers: “For these reasons, tagging data with GUIDs is the single most important step for ensuring the exchange and integration of biological data across the internet.”

Registries and Databases

Local registries. Each laboratory or institution should maintain a registry of specimens containing information about all stored specimens.

Global databases. For data exchange, global databases are used:

• GenBank — nucleotide sequences
• BOLD — barcodes of life
• GBIF — biodiversity data
• Tropicos — botanical nomenclature

As noted by Hyam (2015), “at present GBIF contains 140 million occurrence records linked to names, GenBank contains about 100 million sequences for more than 260,000 named organisms.”

Definition and Structure

A dichotomous key is the most traditional and widespread identification tool. Sharma (2009) defines a key as “an artificial device or analytical means by which a choice is provided between two contradictory statements, leading to the acceptance of one and rejection of the other.”

Structurally, a dichotomous key consists of a sequence of couplets (pairs of mutually exclusive statements). Each statement in a pair is called a lead. The user successively chooses one of two variants until reaching the name of a taxon.

Pandey (2021) emphasizes that “dichotomous keys always have a flow-chart structure, and they can be written either in indented or bracketed form.” Let us consider both forms.

Yoked (Indented) Keys

In a yoked key, each pair of couplets is indented a fixed distance from the left margin. This creates a visual hierarchy that facilitates navigation. Example from Sharma (2009) for five genera of the family Ranunculaceae:

1. Fruit a group of achenes; flowers without spurs

1.1. Petals absent

• Sepals usually 4; involucre absent → Clematis
• Sepals usually 5; involucre present → Anemone
• 2. Petals present → Ranunculus

2. Fruit a group of follicles; flowers with spurs

2.1. Spurs 5; flowers regular → Aquilegia

2.2. Spur 1; flowers irregular → Delphinium

Advantages of yoked keys:

• Visual clarity: the visual structure reflects the logic of the key
• Ease of navigation: the user sees at what level of the hierarchy he or she is located

Limitations:

• Take up much space (each level requires indentation)
• With a large number of taxa, the key becomes cumbersome

Bracketed Keys

In a bracketed key, couplets are arranged sequentially, each leading to the next number. Example from Sharma (2009) for the same five genera:

(1) Fruit a group of achenes; flowers without spurs → (2)

(1) Fruit a group of follicles; flowers with spurs → (4)

(2) Petals absent → (3)

(2) Petals present → Ranunculus

(3) Sepals usually 4; involucre absent → Clematis

(3) Sepals usually 5; involucre present → Anemone

(4) Spurs 5; flowers regular → Aquilegia

(4) Spur 1; flowers irregular → Delphinium

Advantages of bracketed keys:

• Compactness: take up less space
• Standardization: widely used in modern publications

Limitations:

• Less visual for beginners
• Require careful tracking of numbers

Pandey (2021) notes that “in a bracketed key the contrasting characters in a couplet are placed together. The two alternative leads of the couplet are written on adjacent lines, giving a reference by numbers to their successive subordinate leads.”

Principles of Constructing Dichotomous Keys

When constructing keys, a number of principles must be observed to ensure their effectiveness. Sharma (2009) formulates the following recommendations:

1. Accuracy of definitions. “When constructing a key, keep in mind that characters must be precisely defined; measurements should be used wherever possible (do not use terms such as large and small).”

2. Constancy of characters. “Characters that are constant within a taxon are more useful than those that vary.”

3. Availability of characters. “Characters that are available throughout the growing season or easily observed are preferable to those that are ephemeral or difficult to access.”

4. Parallelism of formulations. “Couplets should begin with a noun followed by adjectives, and both leads of a couplet should begin with the same word.”

5. Positive formulations. “Do not use unqualified negative statements; instead use positive statements wherever possible (for example, ‘leaf apex acute’ versus ‘leaf apex pointed’ is better than ‘leaf apex acute’ versus ‘leaf apex not acute’).”

6. Differentiating power. “Keys are most effective when the main couplets divide the entire set of included taxa into more or less equal-sized groups.”

7. Optimal number of characters. “The user is less likely to err if two or three characters are used per lead instead of one, but excessively long leads lead to more errors.”

Advantages and Limitations of Dichotomous Keys

Advantages:

• Simplicity of use: the user only needs to answer questions sequentially
• Clear logical structure: each step leads to an unambiguous result
• Wide distribution: dichotomous keys are available for most groups of organisms
• Possibility of use without special equipment

Limitations:

• Dependence on completeness of material. If a character used in the key is absent or damaged in the specimen, identification becomes impossible. Sharma (2009) notes: “The main disadvantage is that if a particular character is unclear in the specimen for identification, the only available option is to make an assumption about that particular feature.”
• Rigid sequence. The user cannot skip a step or change the order of characters.
• Limited applicability for incomplete material. Dichotomous keys work poorly with fragmentary specimens (for example, only leaves without flowers).
• Difficulty for large groups. With a large number of taxa, the key becomes cumbersome and difficult to use.
• Subjectivity of interpretation. Some characters may be ambiguous, and different users may interpret them differently.

As emphasized by Pandey (2021), “dichotomous keys are widely used for identification of taxa. However, the main disadvantage is that if a particular character is unclear in the specimen for identification, the only available option is to make an assumption about that particular feature.”

Definition and Principle

A polythetic key (also called a multi-access key, polyclave, synoptic key) is an alternative to a dichotomous key, allowing the user to choose characters in any order from a set of available ones.

Pandey (2021) defines a polythetic key as “a list of character states from which the user selects all states corresponding to the specimen and then passes the list to a computer program that matches the list of characters with a database of taxa and characters and returns the identification to the researcher based on matches.”

Sharma (2009) adds that “in the polythetic method of identification the user of the system is free to choose any character, in any order or sequence, and thus avoid the rigid format of traditional dichotomous keys.”

Types of Polythetic Keys

Edge-punched cards. Each card corresponds to a taxon, and holes along the edges correspond to characters. If a character is present in a taxon, the hole is notched. For identification, cards are stacked, and a needle is passed through the hole corresponding to the selected character. Cards with a notched hole fall out; cards without a notch remain on the needle.

Pandey (2021) describes this procedure: “For identification the cards are aligned properly, and a sorting needle is passed through the hole corresponding to a particular character. When the cards are slightly lifted by the needle, the notched cards fall, while the unnotched ones remain on the needle. Thus, this separates the notched cards (i.e., taxa possessing this particular character) from the unnotched cards.”

Body-punched cards. Each card corresponds to a character, and holes correspond to taxa. If a taxon possesses the character, the hole is notched. For identification, cards corresponding to characters present in the specimen are selected; the cards are stacked and viewed against light; the matching hole indicates the taxon.

Pandey (2021) explains: “All selected cards are held against the light, and they are matched in such a way that only one hole is visible. This hole represents the identity of the unknown taxon.”

Computer polythetic keys. Implemented as software (for example, DELTA, Lucid). The user enters available characters, the program matches them with a database, and produces a list of possible taxa.

Sharma (2009) notes that “computerized identification methods include four main approaches: (1) computer-generated keys, (2) computer-stored dichotomous keys, (3) automated pattern recognition systems, and (4) methods of simultaneous character set.”

Advantages of Polythetic Keys

• Flexibility. The user can use any available characters in any order.
• Work with incomplete material. If the specimen lacks some characters, this does not block identification.
• Possibility of obtaining a list of candidates. When information is insufficient, one can obtain a list of possible taxa rather than a single answer.
• Convenience for complex groups. Polythetic keys are especially useful for groups with high variability or for incomplete specimens.

Sharma (2009) emphasizes: “Polythetic keys are especially advantageous when the plant material to be identified is incomplete. Even if the information about the specimen is insufficient to identify a single taxon, the user can obtain a short list of possible identifications for the unknown plant. Thus, polythetic keys can be used as partial keys.”

Limitations of Polythetic Keys

• Labor-intensive preparation. Creating a polythetic key requires more time and effort than a dichotomous one.
• Need for equipment. Edge-punched cards require a physical medium; computer keys require access to technology.
• Less widespread. There are significantly fewer polythetic keys than dichotomous ones.
• Risk of losing cards. Pandey (2021) notes: “The main disadvantage is that if even one card is lost, the entire identification process may go wrong. Also, these cards are subject to wear with each use.”

Vegetative Characters

Vegetative characters are characters associated with vegetative organs (roots, stems, leaves). They are available throughout the growing season and are often used for preliminary identification.

For plants, Sharma (2009) lists the following vegetative characters that should be noted before identification:

• “Determine the herbaceous or woody, annual or perennial nature of the specimen”
• “Note the leaf type, phyllotaxis, and venation”
• “Note the presence or absence and type of stipules on young shoots”
• “Is there milky or colored sap in the leaf, stem, or other parts of the plant”
• “Note the distribution and types of surface coverings (i.e., hairs, trichomes, spines, etc.)”

For insects, vegetative characters are not applicable, but other morphological characters are used: structure of the mouthparts, wing venation, structure of antennae and legs.

Generative Characters

Generative characters are characters associated with reproductive organs (flowers, fruits, seeds; in insects — genitalia). They are often more reliable for identification, since they are less subject to environmental factors.

For plants, Sharma (2009) recommends:

• “Observe the flower and name its parts”
• “Count the number of sepals and petals, and also note their arrangement, i.e., aestivation”
• “Note whether the perianth is present in one series, several series, or absent”
• “Observe whether the sepals and petals are separate or fused”
• “Note whether a pappus (e.g., Asteraceae) or epicalyx (e.g., Malvaceae) or similar structures are present”
• “Determine whether the flowers are actinomorphic or zygomorphic”
• “Observe the number and attachment of stamens. Also note whether there is fusion of anthers or filaments”
• “Count the number of pistils, styles, and stigmas of the gynoecium”
• “Make a cross section of the ovary, count the number of locules, the number of ovules in the locule, and also note placentation”

For insects, generative characters include the structure of genitalia, which are often the only reliable character for distinguishing closely related species. In the IPM manual (Albajes and Madeira, 2019), it is noted that “identification of insects and mites is usually carried out using taxonomic keys,” and genitalia are often used as a key diagnostic character.

Anatomical Characters

Anatomical characters are characters of internal structure that require microscopic analysis.

For plants, anatomical characters include stem structure (arrangement of vascular bundles), leaf structure (type of stomata, mesophyll structure), and root structure. Sharma (2009) recommends “make a cross section of the ovary” to determine the number of locules and type of placentation.

For fungi, anatomical characters include mycelium structure, type of septa, and structure of fruiting bodies. Garrido et al. (2016) describe various staining methods for visualizing pathogen structures: “Resting spores can be distinguished from root tissue using the triple staining method; resting spores stain blue, whereas host tissue stains pink or purple.”

For insects, anatomical characters are used less often, but can be useful for studying internal structures.

Biochemical Characters

Biochemical characters are characters based on the chemical composition of the organism.

For plants, chemotaxonomic characters include composition of flavonoids, alkaloids, and essential oils. These characters are used in systematics at all levels.

For bacteria, biochemical tests are the basis of identification: tests for enzymatic activity, ability to assimilate various substrates, serological reactions.

For fungi, biochemical tests include ability to assimilate various carbon and nitrogen sources and enzyme production.

As noted in the review on pathogen diagnostics (Cubero et al., 2024), “serological methods have not always been replaced by nucleic acid-based technologies, especially when antisera, antibodies, and specific tests are available, sufficiently accurate, and sometimes already commercialized.”

Specificity of Characters for Different Groups of Organisms

Plants. Main diagnostic characters: flower structure (number and arrangement of parts), inflorescence type, fruit structure, leaf shape and venation, type of pubescence, presence and type of stipules.

Insects. Main diagnostic characters: wing structure (venation), mouthparts, antennae, legs, genitalia. In the IPM manual (Albajes and Madeira, 2019), a table of the main orders and families of insect pests is provided with indication of their diagnostic characters.

Fungi. Main diagnostic characters: type of lesion, color and texture of mycelium, shape and size of spores, structure of conidiophores, presence and type of septa.

Bacteria. Main diagnostic characters: cell shape, Gram staining, type of flagellation, biochemical properties, serological reactions.

Variability

Intraspecific variability. The same species may demonstrate significant morphological differences depending on growing conditions, age, and developmental phase. As noted by Sharma (2009), “characters that vary within a taxon are less useful for identification than those that are constant.”

Ecological variability. Environmental conditions can substantially influence morphology. For example, plants of the same species in shade and in open sun may differ significantly in leaf size and shape.

Age variability. Young and adult individuals may differ so much that they are mistaken for different species. This is especially characteristic of insects with complete metamorphosis, where larvae and adults are morphologically completely different.

Polymorphism

Sexual dimorphism. In many organisms, males and females differ morphologically. For example, in some insects males and females may differ in size, color, and structure of antennae.

Seasonal polymorphism. Some species have different morphological forms depending on the season. For example, in many butterflies the spring and summer generations may look like different species.

Caste polymorphism. In social insects (ants, termites, bees), different castes (workers, soldiers, queens) differ morphologically.

Cryptic Species

Cryptic species are species that are morphologically indistinguishable but genetically isolated. Letsiou et al. (2023) define them as “two or more species that are classified as one because they are superficially morphologically similar and therefore cannot be identified using classical taxonomy.”

The problem of cryptic species is especially acute in the agrosphere, where different species may differ in resistance to pesticides, host range, and harmfulness. As noted in the review on DNA barcoding (Letsiou et al., 2023), “the greatest cryptic diversity to date has been found in animals, especially invertebrates.”

Hybrids

Hybrids between closely related species may combine characters of both parents, making their identification difficult. Hybrids may be fertile or sterile, which influences their distribution and significance.

As noted in the manual on systematics (Judd et al., 2018), “hybridization followed by polyploidy can also lead to complex patterns of gene trees.” For identification of hybrids, molecular methods are often required.

Incomplete Material

The absence of diagnostic characters (for example, flowers in a plant or genitalia in an insect) makes morphological identification impossible or unreliable. Sharma (2009) emphasizes that “do not collect specimens that lack flowers and fruits; such material will be difficult to identify.”

Subjectivity of Interpretation

Some morphological characters may be ambiguous, and different specialists may interpret them differently. This is especially characteristic of quantitative characters (sizes, proportions), which may overlap between species.

Keys for Plants

“Flora”. Traditional “Floras” are comprehensive descriptions of plants of a certain territory, including keys, descriptions, illustrations, and information on distribution. Watson et al. (2015) describe the history of the development of Floras: “Frodin (2001) regards Flore francaise by J. B. de Lamarck and A. P. de Candolle (1805–1815) as the first modern Flora. This pioneering work was, in essence, the third edition of Lamarck’s Flore francoise (1778), a publication that is highly significant in the historical discussion of Floras for two main reasons: it was the first floristic work containing Lamarck’s analytical dichotomous keys for species identification.”

Modern Floras. Modern Floras are often available electronically. Watson et al. (2015) note: “For the FAPSA project, the Padme database is crucial for fulfilling another main role of floristic projects — providing convenient, accessible identification tools for plants of a designated region. Padme is critically important for this on several levels. Fundamentally, by storing more detailed, encyclopedic taxonomic information on a web platform, it allows the printed Flora to become increasingly concise and focused on identification of plant species.”

Examples of keys:

• “Flora of North America” — comprehensive Flora of North America
• “Flora of China” — Flora of China
• “Flora Europaea” — Flora of Europe
• “Identification key to plants of Central Russia” — a regional key

Keys for Insects

Dichotomous keys for insects. In the IPM manual (Albajes and Madeira, 2019), a table of the main orders and families of insect pests is provided with indication of their diagnostic characters. For example:

• Order Hemiptera, family Pentatomidae (stink bugs): “They suck plant sap from various tissues, causing wilting of plants, abortion of fruits, or deformation of tissues”
• Order Lepidoptera, family Noctuidae (owlet moths): “Larvae feed on leaves, stems, and fruits, devastating many crops”
• Order Coleoptera, family Chrysomelidae (leaf beetles): “Adults and larvae feed on foliage and fruits; in some other species larvae feed on roots”

Illustrated keys. For insects, illustrated keys are often used, since visual recognition plays a key role.

Keys for Fungi

Mycological keys. For fungi, specialized keys based on microscopic characters are used. Garrido et al. (2016) describe various staining methods for visualizing pathogen structures.

Molecular methods in mycology. For many groups of fungi, morphological identification is difficult, and molecular methods are acquiring increasing importance. As noted in the review on pathogen diagnostics (Cubero et al., 2024), “molecular methods allow more accurate characterization and differentiation of phytopathogenic fungi.”

Keys for Bacteria

Biochemical tests. For bacteria, biochemical tests are traditionally used (tests for enzymatic activity, ability to assimilate various substrates).

Serological methods. Serological methods (ELISA, immunofluorescence) are widely used for diagnosing bacterial pathogens.

Molecular methods. PCR and sequencing are becoming the standard for bacterial identification. As noted in the review on pathogen diagnostics (Cubero et al., 2024), “nucleic acid-based methods proliferated primarily because of their advantages in terms of sensitivity and specificity.”

Polymerase Chain Reaction (PCR)

Polymerase chain reaction (PCR) is a method of amplifying (multiplying) specific regions of DNA. PCR allows obtaining millions of copies of a target fragment from a minimal amount of starting material.

Principle of PCR. The reaction includes three main stages that repeat cyclically:

1. Denaturation — separation of double-stranded DNA into single strands at high temperature (usually 94–95 °C).

2. Annealing — attachment of primers (short synthetic oligonucleotides) to complementary DNA regions at reduced temperature (usually 50–65 °C).

3. Extension — synthesis of a new DNA strand by the enzyme DNA polymerase at intermediate temperature (usually 72 °C).

Each cycle doubles the amount of target DNA, which allows obtaining millions of copies in 30–40 cycles.

Components of PCR:

• Template DNA
• Forward and reverse primers
• DNA polymerase (usually Taq polymerase)
• Deoxynucleoside triphosphates (dNTPs)
• Buffer solution
• Magnesium ions (Mg²⁺)

As noted in the manual on metabarcoding (Lentendu et al., 2022), “PCR reactions are carried out in 96-well plates in a volume of 20 μl, consisting of 1.2 μl of each primer (5 μM), 4 μl buffer, 2 μl 25 mM MgCl₂, 0.6 μl 10 mM dNTPs, 0.1 μl 5 U/μl Taq polymerase, 9.9 μl PCR-clean H₂O, and 1 μl 5–10 ng/μl template DNA.”

Sequencing

Sequencing is the determination of the nucleotide sequence of DNA. There are several sequencing methods, but the most widespread until recently was the Sanger method (dideoxy method).

Principle of the Sanger method. Dideoxynucleoside triphosphates (ddNTPs) are added to the reaction mixture, which terminate chain elongation. As a result, a set of fragments of different lengths is formed, each ending with a certain nucleotide. Separation of fragments by length (for example, by capillary electrophoresis) allows determining the sequence.

Modern methods. In the last two decades, the development of next-generation sequencing (NGS) technologies has revolutionized molecular biology. As noted in the review on pathogen diagnostics (Cubero et al., 2024), “in less than 25 years since the beginning of the ‘omics era’ with the first genome sequence of a free-living phytopathogen, the bacterium Xylella fastidiosa (Simpson et al., 2000), a revolution has occurred in NGS and its application in understanding the molecular bases of pathogen and host biology.”

Advantages of NGS:

• High throughput (millions of sequences per run)
• Reduced sequencing cost
• Possibility of analyzing metagenomes (communities of organisms)

Limitations of NGS:

• Complexity of bioinformatic analysis
• Need for specialized equipment
• Shorter read lengths (for some platforms)

Stages of Molecular Identification

The process of molecular identification includes the following stages:

1. DNA extraction — isolation of DNA from the sample. The quality and purity of DNA are critically important for the success of subsequent stages.

Pandey (2021) notes that “DNA barcoding includes extraction of genomic DNA from tissue samples collected from an individual organism. This is the first step, followed by amplification of the desired region using polymerase chain reaction.”

2. Amplification of the target region — PCR with specific primers.

3. Sequencing — determination of the nucleotide sequence of the amplicon.

4. Bioinformatic analysis — comparison of the obtained sequence with reference databases.

Pandey (2021) emphasizes that “bioinformatics tools are used to analyze the obtained sequences.”

Choice of Primers

The choice of primers is a critical stage of molecular identification. Primers must be:

• Specific — amplify only the target region
• Universal (for a broad range of taxa) or specific (for a particular species)
• Efficient — ensure stable amplification

As noted in the manual on metabarcoding (Lentendu et al., 2022), “in each case information about primers is most important when starting screening of various reported candidate genes for their suitability. Forward and reverse primers must be carefully selected.”

Definition and Principle

DNA barcoding is a method of species identification based on the use of standardized short DNA sequences. Letsiou et al. (2023) define a DNA barcode as “standardized DNA sequences, ideally unique, coding or non-coding, either from the genome of the organism or from its organelles, 400 to 800 base pairs in length, which are used for identification/classification of a group of organisms.”

Key characteristics of a DNA barcode:

• Standardization — the same region is used for all representatives of a group
• Universality — primers work for a broad range of taxa
• Variability — the sequence differs between species but is relatively constant within a species
• Sufficient length — ensures statistical reliability

Pandey (2021) emphasizes that “an ideal DNA barcode should have adequate conservative regions for designing universal primers, high PCR amplification efficiency, and sufficient variability for use in species identification.”

Markers for Different Groups of Organisms

Animals. For animals, the standard marker is a fragment of the mitochondrial gene cytochrome c oxidase I (COI). Letsiou et al. (2023) note that “a 648 base pair fragment near the 5' end of the mitochondrial gene cytochrome c oxidase subunit I (COI) was selected as the standard barcode.”

Advantages of COI:

• Large number of copies per cell (easier to amplify from degraded samples)
• Maternal inheritance (absence of recombination)
• Rapid accumulation of mutations

Plants. For plants the situation is more complex. Letsiou et al. (2023) emphasize that “in plants the use of a universal DNA barcode such as COI, which is used in animals, has still not been achieved.” COI is unsuitable for plants because of the slow mutation rate.

Proposed markers for plants:

• rbcL (ribulose bisphosphate carboxylase/oxygenase large subunit) — conservative, easily amplified, but insufficiently variable for distinguishing closely related species
• matK (maturase K) — more variable, but more difficult to amplify
• trnH-psbA — the most variable non-coding region, but difficult to align
• ITS (internal transcribed spacer) — nuclear marker, highly variable

CBOL Plant Working Group (2009) proposed the combination rbcL + matK as the core barcode for plants. Letsiou et al. (2023) note that “rbcL provides high universality but lower resolution at the species level, whereas matK provides higher resolution but lower universality. This combination provides the best discrimination between species.”

Additional markers: ITS2, trnL-trnF, psbK-psbI, atpF-atpH.

Fungi. For fungi, the standard marker is ITS (internal transcribed spacer) of nuclear ribosomal DNA. Letsiou et al. (2023) note that “the nuclear ribosomal internal transcribed spacer (ITS) was selected as the universal DNA barcoding marker for fungi.”

DNA Barcoding Procedure

The procedure includes the following stages:

1. Sample collection — observing all requirements for DNA preservation (silica gel, ethanol, freezing).

2. DNA extraction — isolation of total DNA from the sample.

3. PCR amplification — amplification of the target region using universal primers.

4. Sequencing — determination of the amplicon sequence.

5. Comparison with a reference database — search for matches in BOLD, GenBank, and other databases.

Pandey (2021) describes this process: “The DNA barcoding process includes two main stages: (a) creation of a DNA barcode library of known species and (b) matching the barcode sequence of an unknown sample with the barcode library for identification.”

Advantages of DNA Barcoding

• Independence from morphology. It is possible to identify samples lacking diagnostic characters (larvae, sterile plants, fragments).
• Standardization. A single approach for all representatives of a group.
• Possibility of automation. High throughput.
• Data accumulation. Each project contributes to the global library.

Letsiou et al. (2023) emphasize that “DNA is not influenced by external or other environmental factors, it is stable and can be found in all tissues, and thus this method can be very useful.”

Limitations of DNA Barcoding

• Lack of a universal marker for plants. Letsiou et al. (2023) note that “an ideal DNA barcode does not exist in plants.”
• Problem of reference databases. “DNA barcodes cannot be a useful identification tool without a comprehensive and reliable reference database” (Letsiou et al., 2023).
• Hybridization and polyploidy. May lead to ambiguous results.
• Recently diverged species. May have identical or nearly identical barcodes.
• Contamination. Possibility of contamination of the sample with foreign DNA.

Principle of Quantitative PCR

Quantitative real-time PCR (qPCR, real-time PCR) is a method that allows not only detecting but also quantifying the content of target DNA in a sample.

Principle. A fluorescent dye or probe that binds to amplicons is added to the reaction mixture. Fluorescence intensity is measured in real time and is proportional to the amount of amplicons.

qPCR chemistries:

• Intercalating dyes (SYBR Green I) — bind to any double-stranded DNA
• Hydrolysis probes (TaqMan) — specific to the target sequence
• Molecular beacons — specific oligonucleotide probes
• Scorpions — bifunctional molecules

As noted in the review on pathogen diagnostics (Cubero et al., 2024), “qPCR allows monitoring the reaction during amplification using a fluorescent signal that increases proportionally to the number of amplicons and the number of targets in the sample.”

Application of qPCR for Pathogen Diagnostics

Advantages of qPCR:

• High sensitivity — detection of single copies of target DNA
• Quantitative assessment — determination of pathogen titer
• Rapidity — result within several hours
• Multiplexing — simultaneous detection of several pathogens
• Automation — minimal risk of contamination

Cubero et al. (2024) emphasize that “in recent years conventional PCR has been replaced by real-time qPCR, which usually has higher sensitivity and relies on automated systems in which no further PCR processing is required for visualization, with the consequent advantage of lower risk of laboratory contamination.”

Examples of application:

• Diagnostics of Plasmodiophora brassicae (causal agent of clubroot of crucifers) — Tso et al. (2021) describe the development of qPCR assays for quantifying spores in soil
• Diagnostics of Xylella fastidiosa — Cubero et al. (2024) describe the use of qPCR for detecting the pathogen in plant tissues
• Diagnostics of Verticillium dahliae — Tso et al. (2021) describe a multiplexed TaqMan real-time PCR assay

Problem of Pathogen Viability

One of the key limitations of molecular methods is the inability to distinguish a viable pathogen from a dead one. As noted by Cubero et al. (2024), “molecular methods have a limitation: they detect microorganisms in any physiological state or simply inert traces of DNA or RNA molecules from dead microorganisms.”

Solution: viability PCR. To distinguish viable and nonviable cells, intercalating agents such as propidium monoazide (PMA) or ethidium monoazide (EMA) are used. These substances penetrate through damaged membranes of dead cells and covalently bind to DNA, preventing its amplification.

Cubero et al. (2024) describe this approach: “Strategies include making the nucleic acids from damaged microorganisms non-amplifiable by covalent binding with ethidium monoazide or PMA upon photoactivation. The approach relies on the integrity of a microorganism’s outer membranes, assuming those that are degraded and allow the entry of ethidium monoazide, PMA, or its new improved version PMAxx correspond to nonviable organisms.”

Limitations of viability PCR:

• Incomplete suppression of amplification of dead cells
• Possibility of false-positive results at high target concentrations

Digital PCR (ddPCR)

Digital PCR (droplet digital PCR, ddPCR) is a method in which a sample is divided into thousands of water-in-oil droplets, each containing zero or one copy of the template. Amplification occurs independently in each droplet, and the number of positive droplets allows absolute quantification of the target DNA content.

Advantages of ddPCR:

• Absolute quantification (without a standard curve)
• High resistance to PCR inhibitors
• High accuracy

Cubero et al. (2024) note that “ddPCR has already been applied for detection of various phytopathogens, including viruses, bacteria, fungi, or oomycetes.”

Definition and Principle

Metabarcoding is a method of analyzing communities of organisms without isolating individual specimens. Lentendu et al. (2022) define it as “a method of identifying multiple species in a single environmental sample using universal PCR primers to simultaneously amplify many DNA barcodes.”

Key characteristics:

• Community analysis — identification of all organisms in a sample
• Universal primers — amplification of a broad range of taxa
• High-throughput sequencing — massively parallel sequencing
• Bioinformatic analysis — processing of large data volumes

Lentendu et al. (2022) emphasize that “metabarcoding is a method of identifying multiple species in a single environmental sample... Unlike traditional DNA barcoding, which attempts to identify intact specimens to species level, metabarcoding aims to identify in various samples or in degraded DNA samples, like environmental samples (eDNA) up to the family level or higher.”

Application of Metabarcoding

Soil communities. Lentendu et al. (2022) describe the application of metabarcoding for analyzing soil eukaryotes: “We describe easy-to-use approaches for studying protists for non-experts... We propose here this approach in order to reveal the widest possible range of soil eukaryotes.”

Plant microbiome. Cubero et al. (2024) note that “metabarcoding, as an amplicon approach, is currently a more widely available approach for applying NGS for diagnostics and surveillance of phytopathogenic prokaryotes, fungi, and oomycetes.”

Biomonitoring. Metabarcoding is used to assess biodiversity in various environments, including water, air, and soil.

Metabarcoding Procedure

The procedure includes the following stages:

1. Sample collection — observing requirements for DNA preservation.

2. DNA extraction — isolation of total DNA from the sample.

3. PCR amplification — amplification of the target region using universal primers.

4. Library preparation — addition of adapters and indices.

5. High-throughput sequencing — massively parallel sequencing.

6. Bioinformatic analysis — filtering, clustering, taxonomic identification.

Lentendu et al. (2022) describe this process: “The protocol is a six-stage process: (1) PCR; (2) sequencing library preparation; (3) NGS; (4) read filtering; (5) sequence assembly; (6) taxon identification.”

Advantages of Metabarcoding

• Analysis of complex communities. Possibility of identifying all organisms in a sample simultaneously.
• Detection of unculturable organisms. Many microorganisms cannot be cultured in the laboratory.
• High throughput. Analysis of hundreds of samples per run.
• Relative quantification. Possibility of estimating proportions of taxa.

Cubero et al. (2024) emphasize that “metabarcoding provides realistic approximations of species abundance when it was used to characterize species of Colletotrichum on walnut.”

Limitations of Metabarcoding

• Relative rather than absolute quantification. Lentendu et al. (2022) note that “metabarcoding cannot provide absolute quantities, only estimates of proportions of pathotypes.”
• Dependence on reference databases. Accuracy of identification depends on the completeness of databases.
• Possibility of contamination. The high sensitivity of NGS makes contamination a serious problem.
• Complexity of bioinformatic analysis. Specialized software and expertise are required.
• Mixed samples. Interpretation of results may be difficult.

Cubero et al. (2024) note that “metabarcoding has several limitations. Bioinformatic expertise and access to bioinformatic facilities are required to analyze sequencing data competently. Sample contamination is of great concern because of the high sensitivity of NGS platforms.”

Contamination

Problem. Contamination of a sample with foreign DNA can lead to false-positive results. This problem is especially acute for metabarcoding, where the high sensitivity of NGS allows detecting even single molecules of a contaminant.

Sources of contamination:

• Reagents and consumables
• Equipment
• Personnel
• Environment

Control measures:

• Use of negative controls
• Work under sterile conditions
• Use of DNA purification kits
• Bioinformatic filtering

Lentendu et al. (2022) emphasize that “proper handling of materials during sampling and nucleic acid extraction, along with the use of environmental controls taken at the sampling site and in the laboratory, helps distinguish the actual set of organisms present in the analyzed sample.”

Low Titer

Problem. When the content of the target organism in a sample is low, molecular methods may give false-negative results.

Causes:

• Insufficient sensitivity of the method
• Presence of PCR inhibitors
• DNA degradation

Solution:

• Increasing the number of PCR cycles (for conventional PCR)
• Use of more sensitive methods (qPCR, ddPCR)
• Optimization of DNA extraction

Cubero et al. (2024) note that “molecular methods have a limitation: they detect microorganisms in any physiological state or simply inert traces of DNA or RNA molecules from dead microorganisms. This dilemma of specific detection of viable organisms may be especially relevant in the case of reproductive or post-harvest materials.”

Mixed Samples

Problem. When analyzing mixed samples (for example, soil, plant material with several pathogens), interpretation of results may be difficult.

Causes:

• Different amplification efficiency for different taxa
• Primer competition
• Uneven distribution of organisms

Solution:

• Use of several markers
• Quantitative assessment (qPCR, ddPCR)
• Bioinformatic analysis

Cubero et al. (2024) note that “metabarcoding depends on preliminary amplification of the target by PCR, which can be a double-edged sword. On the one hand, the approach makes metabarcoding a highly sensitive detection tool, but on the other hand, it carries a significant probability of introducing external contamination, which can be amplified with each PCR cycle, potentially yielding false or low-quality data.”

Lack of Reference Sequences

Problem. The accuracy of molecular identification depends on the completeness of reference databases. If the sequence of the target organism is absent from the database, identification is impossible or unreliable.

Causes:

• Insufficient taxon coverage
• Errors in databases
• Outdated names
• Different taxonomic concepts

Letsiou et al. (2023) emphasize that “DNA barcodes cannot be a useful identification tool without a comprehensive and reliable reference database.”

Hyam (2015) adds that “scientific, type-based names are not adequate identifiers of taxa. Because of this, most names are to some extent ambiguous, but there appears to be little movement toward changing working practice and referring to taxa by reference to specific treatments or using GUIDs.”

Different Taxonomic Concepts

Problem. Different researchers may use different taxonomic concepts, which leads to ambiguity in interpreting results.

Hyam (2015) illustrates this problem: “Take a hypothetical case where an ecologist wants to combine two studies S1 and S2 in a meta-analysis. The studies include taxon names that occur in two taxonomic treatments, T1 and T2. T1 recognizes two taxa under the names A and B, while T2 recognizes only one taxon under the name A and places the name B as a synonym of it. The ecologist must decide whether it is safe to combine data from S1 and S2.”

Problem of Viability

Problem. Molecular methods do not distinguish DNA from living and dead organisms. Detection of pathogen DNA does not necessarily mean the presence of a viable pathogen.

Cubero et al. (2024) note that “detection of traces of nucleic acid from a virus, bacterium, or fungus in treated or disinfected fruits or seeds may have no significance, because the pathogen will not be able to spread from the fruit or, in the case of a seed pathogen, produce a diseased plant.”

Solution: viability PCR using PMA/EMA.

Situations Requiring Molecular Identification

1. Cryptic species. When species are morphologically indistinguishable but genetically isolated.

2. Incomplete material. When the sample lacks diagnostic characters (larvae, sterile plants, fragments).

3. Mixed samples. When it is necessary to identify all components of a community.

4. High accuracy required. For quarantine purposes, forensic examination, breeding programs.

5. Determination of intraspecific categories. Races, pathotypes, biotypes.

6. Rapid diagnostics. When a result must be obtained within tight deadlines.

Cubero et al. (2024) emphasize that “in cases where accurate identification of a specific type of pathogen is required, for example one that requires molecular characterization, very precise diagnostics is of great importance. This approach may make analysis of an extensive sample set unfeasible, thereby de-emphasizing the importance of speed.”

Situations When Molecular Identification Is Excessive

1. Well-known species with clear morphological characters. When morphological identification is reliable and rapid.

2. Preliminary screening. When it is sufficient to determine family or genus.

3. Limited resources. When there is no access to a molecular laboratory.

4. Field conditions. When a quick decision on site is necessary.

General Characteristics

GenBank is the largest public database of nucleotide sequences, maintained by the U.S. National Center for Biotechnology Information. Pandey (2021) defines GenBank as “a repository of a huge amount of publicly available data.”

Main characteristics:

• Contains DNA and RNA sequences from all organisms
• Updated daily
• Includes annotations (descriptions of functions, literature references)
• Integrated with other databases (EMBL, DDBJ) within the International Nucleotide Sequence Database Collaboration

Letsiou et al. (2023) note that “GenBank and the Barcode of Life Data System (BOLD) provide a repository for DNA sequences.”

Structure and Functionality

Record formats:

• GenBank format — text format with annotations
• FASTA format — simplified format (only sequence and header)
• XML format — for programmatic processing

Search:

• By organism name
• By accession number
• By sequence (BLAST)
• By keywords (for example, “barcode,” “ITS,” “rbcL”)

BLAST (Basic Local Alignment Search Tool) is a tool for searching homologous sequences. It allows comparing an unknown sequence with a database and finding the closest matches.

Pandey (2021) describes the procedure: “Bioinformatics tools are used to analyze the obtained sequences.” BLAST is one such tool.

Application in Identification

For molecular identification:

• Search for reference sequences for comparison
• Verification of species affiliation of a sample
• Detection of contamination

For phylogenetic studies:

• Obtaining sequences for phylogenetic analysis
• Comparison with published data

For primer development:

• Search for conservative regions
• Verification of primer specificity

Limitations of GenBank

• Errors in annotations. Sequences may be incorrectly identified.
• Lack of verification. GenBank accepts data from researchers without independent checking.
• Outdated names. Taxonomic names may not correspond to modern concepts.
• Incomplete coverage. For many taxa, reference sequences are absent.

Letsiou et al. (2023) emphasize that “DNA barcodes cannot be a useful identification tool without a comprehensive and reliable reference database.”

General Characteristics

GBIF (Global Biodiversity Information Facility) is an international network and information infrastructure providing open access to biodiversity data.

Main characteristics:

• Contains occurrence records
• Includes information on taxonomy, distribution, and ecology
• Integrates data from many sources (museums, herbaria, observation networks)
• Provides access via web interface and API

Hyam (2015) notes that “GBIF contains 140 million occurrence records linked to names.” This figure is constantly growing.

Data Structure

Occurrence records — records of organism occurrences. Each record includes:

• Taxonomic name
• Geographic coordinates
• Date of observation
• Data source
• Information about the collector

Taxonomic information:

• Taxon name
• Synonyms
• Taxonomic hierarchy

Metadata:

• Information about the dataset
• License
• Citation

Application in Identification

For checking distribution:

• Determination of probable species for a given locality
• Verification of correspondence of a find to the known range

For ecological studies:

• Analysis of species distribution
• Monitoring of range changes

For verification of identification:

• Comparison with known finds
• Identification of unusual finds

Hyam (2015) emphasizes that “GBIF uses a checklist format based on the Darwin Core standard for exchanging taxonomic data.”

Limitations of GBIF

• Errors in identification. Records may contain incorrect identifications.
• Uneven coverage. Data are distributed unevenly across taxa and regions.
• Different taxonomic concepts. Different sources may use different names for the same taxa.
• Lack of verification. Not all records have been checked by specialists.

General Characteristics

BOLD (Barcode of Life Data System) is a specialized database for DNA barcoding, maintained by the University of Guelph (Canada).

Main characteristics:

• Contains DNA barcodes for all groups of organisms
• Includes metadata on voucher specimens
• Provides tools for analysis and identification
• Integrated with GenBank

Pandey (2021) notes that “BOLD provides an integrated bioinformatic platform that supports all phases of the analytical path from specimen collection to a tightly validated barcode library.”

Data Structure

Barcode records — records of DNA barcodes. Each record includes:

• Barcode sequence
• Information about the voucher specimen
• Taxonomic name
• Geographic origin
• Literature references

BIN (Barcode Index Number) — a system of sequence clustering that assigns a unique number to clusters of closely related sequences. This allows using BIN as a “proxy species” for identification.

Application in Identification

For identification of samples:

• Uploading a sequence and searching for matches
• Obtaining a list of probable taxa
• Assessment of identification reliability

For creating reference libraries:

• Uploading one’s own data
• Comparison with existing records

For data quality checking:

• Identification of errors in identification
• Detection of contamination

Letsiou et al. (2023) emphasize that “BOLD provides an integrated bioinformatic platform that supports all phases of the analytical path from specimen collection to a tightly validated barcode library.”

Limitations of BOLD

• Incomplete coverage. For many taxa, reference barcodes are absent.
• Errors in identification. As in GenBank, data may contain errors.
• Dependence on morphological verification. BOLD requires that specimens be identified morphologically.
• Limited applicability for plants. For plants, several markers are used, which complicates standardization.

General Characteristics

Tropicos is a botanical database maintained by the Missouri Botanical Garden. It contains nomenclatural, taxonomic, and bibliographic information about plants.

Main characteristics:

• Contains information on more than 1.3 million plant names
• Includes data on type specimens
• Provides access to images of herbarium specimens
• Integrated with other botanical resources

Data Structure

Nomenclatural records:

• Taxon name
• Author of the name
• Place of publication
• Type specimen

Taxonomic records:

• Accepted name
• Synonyms
• Taxonomic hierarchy

Bibliographic records:

• Literature references
• Protologues

Images:

• Photographs of herbarium specimens
• Images of types

Application in Identification

For checking nomenclature:

• Checking correct spelling of names
• Searching for synonyms
• Determining authors of names

For access to type specimens:

• Viewing images of types
• Obtaining information about the location of types

For botanical research:

• Obtaining information on distribution
• Analysis of nomenclatural history

Sharma (2009) emphasizes the importance of type specimens: “Type specimens are among the most valuable collections of any herbarium, since these are the specimens to which the name of a taxon is permanently attached, and they are useful for confirming the identity of any element.”

Limitations of Tropicos

• Limited coverage. Not all taxa are represented equally fully.
• Outdated data. Some records may not reflect modern taxonomic concepts.
• Language barrier. The interface and descriptions are predominantly in English.
• Lack of molecular data. Tropicos focuses on nomenclature and morphology.

General Characteristics

In recent years many mobile applications have appeared that use machine learning for automatic identification of plants and animals. Jones (2025) notes that “since the late 2010s there has been an explosion in the number and diversity of applications that have been developed for use on smartphones and that aim to automatically identify any plants that one might come across. A quick Google Play search reveals more than 40 such apps available, many of them free.”

Main applications:

• PlantNet — an application developed by French researchers, using crowdsourcing to collect images
• iNaturalist — a platform for citizen science combining automatic identification with community verification
• Seek — an application from iNaturalist aimed at beginners
• Flora Incognita — an application developed in Germany, focusing on European flora
• PictureThis — a commercial application with broad coverage

Principle of Operation

Machine learning. Applications use neural networks trained on large image databases. As noted by Jones (2025), “these deep learning systems are based on training systems using vast databases of images and using neural networks; crowdsourcing of data allows steady improvement of systems as they are used.”

Identification process:

1. The user photographs the organism

2. The application analyzes the image

3. The neural network compares the image with the training sample

4. The application produces a list of probable identifications with indication of confidence

Effectiveness of Applications

Jones (2025) reports the results of testing applications: “All tested applications improved over the 3 years between 2020 and 2023 by approximately 20 percentage points in terms of the percentage of samples correctly identified to species or family. The best applications, which include Flora Incognita and PlantNet, were found to be able to correctly identify to species 80–90% of the plants tested here.”

Factors influencing accuracy:

• Image quality
• Presence of flowers/fruits
• Group of organisms
• Geographic region

Jones (2025) notes that “most studies did not report significant differences between plant groups (monocots, grasses, shrubs, trees, etc.) in the proportion of successful identifications, although in most studies discriminating power was low due to low sample numbers.”

Advantages of Mobile Applications

• Accessibility. Applications are free or inexpensive, available on smartphones.
• Rapidity. Result within a few seconds.
• Simplicity of use. They do not require special knowledge.
• Education. They help beginners master identification.
• Citizen science. They involve the general public in data collection.

Jones (2025) emphasizes that “the development and wide availability of automatic identification applications have significantly expanded opportunities for the general public to participate in plant identification and citizen science projects.”

Limitations of Mobile Applications

• Errors in identification. Applications may give incorrect results, especially for rare or cryptic species.

Jones (2025) notes that “applications may be biased against identifying uncommon or rare species, partly because of limited training data available for such species.”

• Lack of confidence. Not all applications indicate the level of confidence in the result.

Jones (2025) emphasizes that “many of the other popular applications, including Leafsnap, PictureThis, and PlantSnap, or more general applications such as Google Lens, do not indicate the level of confidence in their results.”

• Failure to account for geographic variability. Applications may not take into account regional features of morphology.

• Cryptic species. Applications cannot distinguish species that are morphologically indistinguishable.

• Dependence on training sample. The accuracy of an application depends on the completeness and quality of training data.

The Rule “AI Does Not Determine Species, AI Proposes a Hypothesis”

This rule, formulated in the title of Lecture 6, is of fundamental importance. AI-based applications do not give a final answer — they propose a hypothesis that requires verification.

Jones (2025) formulates it as follows: “We conclude that the specification for ideal applications is that they should provide an indication of confidence in the results and minimize incorrect identifications, even if it is only wrong species of two similar ones.”

Practical recommendations:

• Always verify the result of an application with other sources
• Use several applications for cross-checking
• Pay attention to the level of confidence
• Consult a specialist for critical identifications

CABI

General characteristics. CABI (Centre for Agriculture and Biosciences International) is a non-profit international organization specializing in agricultural information.

Sharma (2009) notes that “CABI is a non-profit international organization that has particular relevance in the agricultural community because of its goal of improving people’s lives by providing information and applying scientific expertise to solve problems. Transferring knowledge from the laboratory to the field is the mission of this organization, which uses scientific publishing, development projects, research, and microbiological services.”

CABI resources:

• CAB Abstracts — database on agricultural sciences
• Crop Protection Compendium — reference book on plant protection
• Invasive Species Compendium — reference book on invasive species
• Plantwise — program on plant disease diagnostics

EPPO

General characteristics. EPPO (European and Mediterranean Plant Protection Organization) is an intergovernmental organization responsible for cooperation in plant protection in Europe and the Mediterranean region.

Garrido et al. (2016) describe EPPO: “EPPO is an intergovernmental organization with 50 member countries, responsible for cooperation in plant protection in Europe and the Mediterranean region. This organization aims to help its member countries prevent the introduction or spread of dangerous pests.”

EPPO resources:

• EPPO Global Database — database on quarantine organisms
• EPPO Standards — diagnostic standards
• Pest Risk Analysis — pest risk analysis
• EPPO Reporting Service — reports on new findings

EPPO lists:

• A1 List — quarantine organisms absent from the EPPO region
• A2 List — quarantine organisms present in the EPPO region but limited in distribution
• Alert List — organisms representing a potential threat

Garrido et al. (2016) note that “in recent years several genera of fungal pathogens of strawberry, such as Colletotrichum acutatum, Botrytis cinerea, and Phytophthora spp., have been included in the EPPO A2 list.”

Application in Identification

For checking quarantine status:

• Determining whether an organism is quarantine
• Obtaining information on phytosanitary measures

For diagnostics:

• Using EPPO diagnostic standards
• Obtaining information on detection and identification methods

For risk analysis:

• Assessing the probability of introduction and spread
• Determining potential impact

Limitations

• Limited coverage. Not all organisms are included in the databases.
• Updating. Information may lag behind the current situation.
• Access. Some resources require subscription.
• Language barrier. Not all resources are available in all languages.

Errors in Databases

Sources of errors:

• Incorrect identification of specimens
• Data entry errors
• Outdated names
• Contamination of specimens
• Errors in sequences

Consequences:

• False-positive identifications
• False-negative identifications
• Incorrect conclusions

Control measures:

• Data verification
• Cross-checking
• Use of several sources

Letsiou et al. (2023) emphasize that “DNA barcodes cannot be a useful identification tool without a comprehensive and reliable reference database.”

Outdated Names

Problem. Taxonomic names are constantly revised. A name valid 20 years ago may be a synonym today.

Examples:

• Many species described in the 18th–19th centuries have been reclassified
• Molecular data have led to revision of many groups
• Different taxonomic schools use different names

Consequences:

• Confusion when searching for information
• Mismatch between databases
• Difficulties in data exchange

Hyam (2015) notes that “scientific, type-based names are not adequate identifiers of taxa. Because of this, most names are to some extent ambiguous.”

Different Taxonomic Concepts

Problem. Different researchers may use different taxonomic concepts, which leads to ambiguity in interpreting results.

Hyam (2015) illustrates this problem: “Take a hypothetical case where an ecologist wants to combine two studies S1 and S2 in a meta-analysis. The studies include taxon names that occur in two taxonomic treatments, T1 and T2. T1 recognizes two taxa under the names A and B, while T2 recognizes only one taxon under the name A and places the name B as a synonym of it.”

Consequences:

• Impossibility of automatic data merging
• Need for manual checking
• Risk of incorrect conclusions

Solution. Use of globally unique identifiers (GUIDs) for taxa.

Hyam (2015) emphasizes that “for these reasons, tagging data with GUIDs is the single most important step for ensuring the exchange and integration of biological data across the internet.”

Incomplete Coverage

Problem. Many taxa are insufficiently represented in databases.

Causes:

• Lack of research
• Difficulties of collection
• Lack of funding
• Geographic remoteness

Consequences:

• Impossibility of identification
• Unreliable results
• Limited applicability

Letsiou et al. (2023) note that “for plants the use of a universal DNA barcode such as COI, which is used in animals, has still not been achieved.”

Problem of Data Quality

Problem. Not all data in databases have the same quality.

Quality criteria:

• Accuracy of identification
• Completeness of metadata
• Reliability of source
• Verification

Measures to improve quality:

• Verification by specialists
• Use of standards
• Cross-checking
• Feedback

Hyam (2015) emphasizes that “large volumes of information are tagged with taxonomic names and published on the internet... Unfortunately, these methods can only lead to broadening the query. The process leads to the broadest possible interpretation of any taxon. It adopts a sensu lato approach to everything — the ultimate ‘lumpers’ approach to taxonomy.”

Use of Several Sources

To increase the reliability of identification, it is recommended to use several independent sources:

• Several databases (GenBank, BOLD, GBIF)
• Several applications (PlantNet, iNaturalist, Flora Incognita)
• Several identification keys
• Consultation with a specialist

Critical Evaluation of Data

When working with databases it is necessary to critically evaluate:

• Data source. Who provided the data? Is there verification?
• Date. When were the data added? Are they outdated?
• Completeness. Are there metadata? Is there enough information?
• Consistency. Do data from different sources agree?
• Confidence level. Is the level of confidence indicated?

Documentation

When using databases it is necessary to document:

• Which databases were used
• Which queries were performed
• Which results were obtained
• Which verification measures were applied

Data Exchange

A specialist must not only use databases but also contribute:

• Upload verified sequences to GenBank/BOLD
• Provide occurrence data to GBIF
• Report errors in databases
• Publish data in open repositories

Hyam (2015) emphasizes that “free access to data and their exchange are key to the progress of systematics.”

General Principle

Modern applications for automatic identification of organisms are based on machine learning methods, in particular deep neural networks. Jones (2025) describes this as follows: “Starting in 2013, various approaches and identification algorithms were compared through the LifeCLEF challenge and its PlantCLEF sub-challenge. The approach used generally included specific feature extraction. Around the same time, more general ‘deep learning’ algorithms were being developed, such as the one used in Google Lens. These deep learning systems are based on training systems using vast databases of images and using neural networks; crowdsourcing of data allows steady improvement of systems as they are used.”

Key components:

1. Training sample — a set of images with known identifications

2. Neural network — a mathematical model consisting of layers of “neurons”

3. Training process — adjustment of network parameters to minimize errors

4. Inference — application of the trained network to new images

Architecture of Neural Networks

Modern identification systems use convolutional neural networks (CNNs), which are specially designed for image analysis.

Barhate et al. (2024) describe various architectures used for plant identification:

• AlexNet — one of the first deep CNNs, which showed outstanding results
• VGG-16/VGG-19 — networks with increased depth
• ResNet — networks with “residual connections,” allowing training of very deep models
• Inception — networks with multi-scale processing
• MobileNet — lightweight networks for mobile devices
• DenseNet — networks with dense connections between layers

Barhate et al. (2024) note that “all these architectures were used to extract features from images of leaves, flowers, and other plant organs.”

Training Process

Stage 1: Data collection. A training sample is formed — a large number of images of organisms with known identifications. The larger and more diverse the sample, the better the system will work.

Stage 2: Preprocessing. Images are brought to a unified format, normalized, and may be augmented (rotations, reflections, brightness changes) to increase diversity.

Stage 3: Training. The neural network is “shown” images and their corresponding labels. The network gradually adjusts its parameters to minimize the error between prediction and true label.

Stage 4: Validation. The trained network is tested on a separate sample to evaluate its ability to generalize.

Stage 5: Inference. The trained network is applied to new images.

Crowdsourcing and Continuous Learning

One of the key features of modern applications is the use of crowdsourcing for continuous improvement. Jones (2025) notes that “crowdsourcing of data allows steady improvement of systems as they are used.”

Mechanisms of crowdsourcing:

• Users upload images with identifications
• The community verifies identifications
• The system learns from verified data
• Accuracy gradually increases

Examples:

• iNaturalist — users upload observations, the community confirms identifications
• PlantNet — users help identify images of other users

Training on Limited Samples

Problem. Neural networks are trained on available data. If the training sample is incomplete or unrepresentative, the system will make errors.

Jones (2025) notes that “applications may be biased against identifying uncommon or rare species, partly because of limited training data available for such species.”

Examples:

• Rare species are represented by few images
• Species from remote regions are insufficiently represented
• Seasonal variations (flowering, fruiting) may be represented unevenly

Consequences:

• The system may “fail to recognize” a rare species
• The system may mistakenly identify a rare species as a closely related common one
• Accuracy for rare species is significantly lower than for common ones

Barhate et al. (2024) emphasize that “increasing the number of images or datasets can lead to deterioration in algorithm performance” — this is a counterintuitive but important effect associated with class imbalance.

Failure to Account for Geographic Variability

Problem. The same species may demonstrate significant morphological differences in different parts of its range. A system trained on images from one region may work poorly in another.

Jones (2025) gives an example: “It was unclear why this low success (compared with the present study) occurred, but possibly it could have been because testing was conducted in the USA and the application was developed in Europe, or possibly the plant types were unrepresentative of the native European flora used in developing Flora Incognita.”

Examples:

• Plants of the same species in shade and sun look different
• Seasonal forms may differ
• Geographic races may have different morphological characters

Cryptic Species

Problem. Cryptic species are morphologically indistinguishable but genetically isolated. No image-based system can distinguish them.

Letsiou et al. (2023) define cryptic species as “two or more species that are classified as one because they are superficially morphologically similar and therefore cannot be identified using classical taxonomy.”

Examples:

• Many species of insect pests have cryptic species
• Fungal pathogens may have cryptic species with different virulence
• Weed plants may include cryptic species

Consequences:

• The system will identify a cryptic species as one of the known ones
• It is impossible to distinguish species with different harmfulness
• Molecular validation is required

Intraspecific Variability

Problem. Even within one species, morphological characters may vary significantly.

Sources of variability:

• Age changes (young and adult individuals)
• Sexual dimorphism
• Seasonal polymorphism
• Ecological plasticity

Examples:

• Larvae and adults of insects are completely different
• Males and females may differ in size and color
• Plants under different growing conditions may look like different species

Problem of “Garbage” Images

Problem. Applications may give erroneous results when analyzing low-quality images or images that do not contain diagnostic characters.

Jones (2025) describes an experiment: “To investigate the robustness of different applications to image quality, we generated a series of deliberately poor images with muddled backgrounds, mixed species, or lack of focus, and tested several of the best identification applications on them.”

Results:

• PlantNet and Flora Incognita showed high robustness
• PictureThis made one error
• iNaturalist made four errors

Conclusion: some applications are robust to image quality, but not all.

Problem of the “Closed World”

Problem. Neural networks are trained on a closed set of classes. If an organism does not belong to any of the known classes, the system will still produce the “nearest” answer.

Examples:

• A new, not yet described species will be identified as a closely related known one
• A hybrid will be identified as one of the parents
• An organism from a remote region will be identified as a closely related species from the training sample

Consequences:

• False identifications
• Impossibility of discovering new species
• Incorrect conclusions

Identification of Rare Species as Common Ones

Jones (2025) notes that “applications may be biased against identifying uncommon or rare species, partly because of limited training data available for such species... but also, possibly, because the critical characters for such species may be cryptic.”

Example: A rare orchid species may be identified as a closely related common species because there were few images of the rare species in the training sample.

Identification of Hybrids as Parental Species

Problem. Hybrids often combine characters of both parents. A system not trained on hybrids may identify them as one of the parental species.

Example: A hybrid between two oak species may be identified as one of the parents.

Identification of Damaged Specimens

Problem. Damaged or incomplete specimens may be identified incorrectly.

Example: A leaf without flowers may be identified incorrectly if the system was trained mainly on images with flowers.

Identification Based on Background

Problem. Neural networks may “remember” the background rather than the organism itself. If training images were taken under certain conditions, the system may use the background as a character.

Example: If all images of species A were taken against grass and species B against stones, the system may identify the species by the background rather than by the organism.

Examples from Practice

Identification of poisonous plants. Jones (2025) gives an example: “Another published comparison of application performance includes a careful study with good replication on 17 toxic species found in America (Otter et al., 2021), where PictureThis correctly identified 10 (59%) and PlantNet 8 (47%) species, while PlantSnap correctly identified only one species.”

This means that even the best applications may err in critically important cases (identification of poisonous plants).

Identification of quarantine organisms. An error in identifying a quarantine organism may have serious consequences. As noted by Cubero et al. (2024), “different bacteria of the genus Xanthomonas that infect citrus cause similar symptoms at the beginning of infectious processes. However, depending on the particular Xanthomonas species, the infection may lead to citrus canker or bacterial spot of citrus. The former is considered a very serious and quarantine disease in many countries, whereas the latter is a disease of lesser significance and is usually not regulated.”

Why AI Cannot Replace a Specialist

Lack of understanding. AI does not “understand” the biology of an organism. It recognizes patterns but does not know why they are important.

Lack of context. AI does not take into account ecological context, geographic origin, or phenology.

Lack of responsibility. AI does not bear responsibility for errors. Responsibility lies with the specialist who made the decision.

Limited training data. AI knows only what it has been taught. It cannot go beyond the training sample.

The Role of the Specialist in Verification

Plausibility check. The specialist checks whether the proposed identification corresponds to the ecological context, geographic distribution, and phenology.

Check of diagnostic characters. The specialist checks whether the diagnostic characters of the specimen correspond to the proposed taxon.

Use of additional sources. The specialist uses identification keys, herbarium specimens, and consultations with colleagues.

Decision-making. The specialist makes the final decision on identification and bears responsibility for it.

The Rule “AI Proposes a Hypothesis”

This rule can be formulated as follows:

1. AI generates a hypothesis — a proposal about a possible identification.

2. The specialist tests the hypothesis — using all available methods.

3. The specialist accepts or rejects the hypothesis — on the basis of critical evaluation.

4. The specialist documents the decision — indicating the method and level of confidence.

Jones (2025) formulates it as follows: “We conclude that the specification for ideal applications is that they should provide an indication of confidence in the results and minimize incorrect identifications, even if it is only wrong species of two similar ones.”

Confidence Levels

Ideal scenario. The application indicates the level of confidence for each proposed identification.

Reality. Not all applications indicate the level of confidence. Jones (2025) notes that “many of the other popular applications, including Leafsnap, PictureThis, and PlantSnap, or more general applications such as Google Lens, do not indicate the level of confidence in their results.”

Recommendation. Use applications that indicate the level of confidence (PlantNet, Flora Incognita, iNaturalist).

Legal Responsibility

In the case of quarantine incidents or the use of pesticides causing damage, responsibility is borne by the specialist who made the decision. Reference to “the application said so” is not an justification.

Economic Responsibility

An error in identification may lead to:

• Crop loss
• Unjustified pesticide costs
• Environmental damage
• Loss of reputation

Ethical Responsibility

The specialist bears ethical responsibility for:

• The accuracy of his or her identifications
• The safety of the methods used
• The conservation of biodiversity
• The dissemination of knowledge

Practical Recommendations

1. Never rely only on AI. Use AI as an additional tool, not as the only source.

2. Verify results. Use several applications, identification keys, and consult specialists.

3. Document. Record which methods were used, which results were obtained, and which verification measures were applied.

4. Improve qualifications. Study morphology, taxonomy, and identification methods.

5. Be honest. If you are not sure about an identification, say so. It is better to admit uncertainty than to give false confidence.

Example 1: Plant Identification

Situation. The PlantNet application identified a plant as Conium maculatum (poison hemlock) with 85% confidence.

Actions of the specialist:

1. Check diagnostic characters: presence of spots on the stem, characteristic leaves, umbels.

2. Check ecological context: poison hemlock grows in wet places, often along roads.

3. Check geographic distribution: the species is distributed in the region.

4. Check phenology: flowering in June–July.

5. If necessary — consultation with a specialist.

Decision. If all checks confirm, the identification is accepted. If there are doubts — additional verification is required.

Example 2: Insect Identification

Situation. The iNaturalist application identified an insect as Leptinotarsa decemlineata (Colorado potato beetle) with 90% confidence.

Actions of the specialist:

1. Check diagnostic characters: yellow stripes on the elytra, characteristic body shape.

2. Check the host: Colorado potato beetle feeds on Solanaceae.

3. Check geographic distribution: the species is distributed in the region.

4. Check phenology: adults appear in spring.

Decision. If all checks confirm, the identification is accepted. If there are doubts — additional verification is required.

Example 3: Pathogen Identification

Situation. The application identified a fungal pathogen as Botrytis cinerea (gray mold) with 75% confidence.

Actions of the specialist:

1. Check symptoms: gray fluffy coating, softening of tissues.

2. Check the host: Botrytis cinerea affects many crops.

3. Check conditions: high humidity, temperature.

4. Conduct microscopic analysis: conidiophores, conidia.

5. If necessary — molecular validation.

Decision. If all checks confirm, the identification is accepted. If there are doubts — additional verification is required.

Development Trends

Improving accuracy. Jones (2025) notes that “all tested applications improved over the 3 years between 2020 and 2023 by approximately 20 percentage points.”

Expanding coverage. Applications are gradually including more species and regions.

Integration with other tools. Applications are integrated with databases, identification keys, and monitoring systems.

Improving interface. Applications are becoming more convenient for users.

Prospects

Automatic identification in the field. In the future, applications may become accurate enough for automatic identification in real time.

Integration with molecular methods. Applications may be integrated with portable sequencers for molecular validation.

Citizen science. Applications will continue to involve the general public in collecting biodiversity data.

Limitations That Will Remain

Cryptic species. No image will allow distinguishing morphologically identical species.

New species. The system cannot identify what it has not been taught.

Hybrids. Complex hybrids may remain unresolved.

Context. AI does not understand ecological and evolutionary context.

Principle of Integration

No single identification method is universal. As we have seen, the morphological approach has limitations when working with cryptic species, incomplete material, and hybrids. Molecular methods, in turn, do not distinguish living and dead organisms, depend on the completeness of reference databases, and may give false-positive results upon contamination. Only an integrated approach combining several independent lines of evidence allows achieving reliable identification.

Cubero et al. (2024) emphasize that “both molecular methods and spectral imaging have their strengths and weaknesses. Molecular methods provide a high degree of reliability that spectral imaging lacks, but they are limited by smaller sampling capabilities. Spectral imaging, on the other hand, can quickly cover large areas, although it still requires improvements to increase accuracy and clearly distinguish biotic and abiotic plant health stressors.”

Components of an Integrated Approach

Morphology. Primary analysis of morphological characters remains the starting point of identification. Sharma (2009) recommends beginning with determining the general habit, leaf type, phyllotaxis, venation, presence and type of pubescence, and structure of flower and fruit.

Molecules. Molecular methods are used when morphology is insufficient: for cryptic species, incomplete material, mixed samples, and determination of intraspecific categories. As noted by Letsiou et al. (2023), “DNA is not influenced by external or other environmental factors, it is stable and can be found in all tissues.”

Ecology. Ecological context is an important component of identification. Knowledge of habitat, host, substrate, and phenology allows discarding obviously unsuitable variants. As noted in the IPM manual (Albajes and Madeira, 2019), “plants damaged by pests often demonstrate changes in absorption and reflection in the visible and near-infrared regions due to reduced chlorophyll content, changes in other pigments, or some other changes in the internal anatomy of the plant.”

Geography. Geographic distribution is another filter for testing identification hypotheses. If the proposed species does not occur in the given region, this is a serious argument against the identification. Hyam (2015) emphasizes that “GBIF contains 140 million occurrence records linked to names,” which allows checking species distributions.

Example of an Integrated Approach

Let us consider an example of identifying a fungal pathogen on strawberry.

Step 1: Morphology. Examination of symptoms: brown spots on leaves with a dark center and reddish border. Microscopy: conidia of a certain shape and size.

Step 2: Ecology. Crop — strawberry. Conditions — high humidity, moderate temperature. Season — late summer.

Step 3: Geography. Region — southern Russia. Check via GBIF: which species occur in this region.

Step 4: Molecules (if necessary). PCR with species-specific primers. Sequencing of the ITS region. Comparison with GenBank/BOLD.

Step 5: Synthesis. If all lines of evidence agree, the identification is accepted. If there are contradictions — additional analysis is required.

As noted by Garrido et al. (2016), “molecular methods allow more accurate characterization and differentiation of phytopathogenic fungi.” However, they emphasize that “classical identification methods based on morphology remain important for initial diagnostics.”

Significance of a Reference Collection

A reference collection is a collection of verified specimens that serves as a standard for comparison during identification. As emphasized by Sharma (2009), “herbaria are standard reference centers for identification and verification of new plant material and provide the basis for taxonomic research.”

For an agronomist, a reference collection may include:

• Herbarium specimens of cultivated plants and weeds
• Entomological collections of pests and entomophages
• Collections of affected organs with identified pathogens
• Fixed specimens for molecular analysis

Principles of Forming a Collection

Representativeness. The collection should represent the diversity of organisms with which you work. Judd et al. (2018) emphasize: “Collected plants should represent the variation observed in the population.”

Verification. Each specimen must be identified by a specialist or using reliable methods. Sharma (2009) notes that “only experts of the particular family or genus should be allowed to annotate a herbarium sheet.”

Completeness. The specimen must contain all diagnostic characters. Judd et al. (2018) recommend: “If possible, the whole plant should be collected, including underground parts; never collect only one flower or leaf!”

Documentation. Each specimen must be supplied with a label containing complete information. Sharma (2009) lists the necessary data: scientific name, family, place of collection, altitude above sea level, latitude and longitude, date of collection, collector, collection number, field notes.

Preservation Methods

Herbarization of plants. Judd et al. (2018) describe the process: “The plant is placed in folded newspaper and arranged so that the upper and lower surfaces of leaves, flowers, and fruits are visible.” Drying is carried out in a press with daily changing of blotters.

Preservation of insects. For Coleoptera and Lepidoptera — mounting on pins. For soft-bodied forms (aphids, larvae, mites) — preservation in 70–96% ethanol.

Preservation of fungi. For phytopathogenic fungi — herbarization of affected organs or preservation of pure cultures on nutrient media. Garrido et al. (2016) note that “commercial kits are available for extracting fungal DNA from environmental samples such as plant tissue or soil.”

Preservation for molecular analysis. Lentendu et al. (2022) recommend: “If samples are to be transported more than 2 days before reaching the laboratory, sieve on site or at base camp and store aliquots of 0.5–1 g of soil in at least two 2 ml tubes... and add 1 ml of DNA preservation solution to each tube.”

Organization of Storage

Systematization. The collection must be organized according to the taxonomic principle. Sharma (2009) notes that “the arrangement of specimens in a herbarium usually follows a well-accepted classification system... This allows easy comparison of related taxa, since they are placed close to each other.”

Storage conditions. Sharma (2009) emphasizes: “Herbarium specimens will be preserved indefinitely with careful handling. However, they are susceptible to damage by insects, fungi, and fire, and should be stored in specialized cabinets in a climate-controlled environment.”

Protection from pests. Sharma (2009) describes methods: “Freezing is a preservation method that is least dangerous to human health and quite simple. In this method, specimens are frozen to −18 °C or colder and kept for at least 48 hours.”

When Cooperation Is Necessary

It is not always possible or advisable to carry out identification independently. There are situations when it is necessary to turn to specialists:

• Complex taxonomic groups. Some groups require special knowledge and experience.
• Critical identifications. For quarantine purposes, forensic examination, breeding programs.
• Lack of reference materials. If your collection does not contain the necessary standards.
• Need for molecular analysis. If sequencing or qPCR is required.

As noted by Sharma (2009), “problem collections can be sent to an expert on the systematics of a particular genus or family or on the flora of a particular region.”

Forms of Cooperation

Consultations. Turning to a specialist to check identification. Sharma (2009) notes that “expert identification is one of the most reliable methods of plant identification by experts on a particular taxonomic group.”

Joint research. Participation in projects on inventory, monitoring, and biodiversity study.

Exchange of materials. Providing specimens for research, obtaining reference materials.

Training. Internships, participation in workshops and advanced training courses.

Diagnostic Laboratories

In the agrosphere there are specialized diagnostic laboratories that carry out identification of pathogens, pests, and weeds. Cubero et al. (2024) emphasize that “accurate detection and identification of a pathogen are of great importance,” and in many countries there are official laboratories accredited for such studies.

Functions of diagnostic laboratories:

• Conducting standard diagnostic tests
• Developing new diagnostic methods
• Validating results
• Training specialists

Interaction with laboratories:

• Providing samples
• Obtaining results
• Consultations on methods
• Participation in interlaboratory comparisons

International Cooperation

For quarantine organisms and invasive species, international cooperation is often required. Garrido et al. (2016) describe the role of EPPO: “EPPO is an intergovernmental organization with 50 member countries, responsible for cooperation in plant protection in Europe and the Mediterranean region. This organization aims to help its member countries prevent the introduction or spread of dangerous pests.”

Significance of Documentation

Documentation is not a bureaucratic formality but a necessary element of scientific and practical activity. As emphasized by Hyam (2015), “tagging data with GUIDs is the single most important step for ensuring the exchange and integration of biological data across the internet.”

Documentation allows:

• Verifying results
• Exchanging data
• Tracking changes
• Teaching others

What to Document

For each specimen:

• Taxonomic identification (name, author, identifier)
• Date and place of collection
• Ecological data (habitat, host, substrate)
• Method of collection and fixation
• Relation to other specimens
• Molecular data (accession numbers)

Sharma (2009) emphasizes that “field notes include additional information about the plant that may disappear after drying plants and cannot be understood simply by looking at a herbarium sheet, such as flower color; height, pubescence, duration, or habit of the plant; associated plant species; habitat; branching pattern; population size; vegetation type.”

For each identification:

• Identification method
• Identification keys and databases used
• Level of confidence
• Alternative variants

Data Formats

Standards. Standardized formats are used for data exchange:

• Darwin Core — standard for biodiversity data
• Taxon Concept Schema — for taxonomic data
• GenBank format — for nucleotide sequences

Hyam (2015) notes that “GBIF uses a checklist format based on the Darwin Core standard for exchanging taxonomic data.”

Globally Unique Identifiers (GUIDs). Hyam (2015) emphasizes the importance of GUIDs: “If each record in a dataset has an immutable identifier, then the consumer will immediately see which records have changed, which are additions, and whether the changes may be for biological reasons.”

Data Exchange

Publication in open databases. A specialist must not only use databases but also contribute:

• Uploading verified sequences to GenBank/BOLD
• Providing occurrence data to GBIF
• Publishing images in open repositories

Compliance with licenses. When exchanging data, license restrictions must be observed. Some data may be confidential (for example, exact coordinates of rare species).

Citation. When using data from databases, the source must be cited. Hyam (2015) notes that “the consumer could also individually credit the providers for the use of their data.”

Need for Continuous Learning

Identification is a competence that requires constant improvement. As emphasized by Watson et al. (2015), “in programmes across the Arabian region... there is great potential for using digitally produced plant profiles and interactive identification guides as a major resource for a series of training programmes aimed at building capacity in plant taxonomy (particularly plant identification), for taxonomy students, ecologists, environmental impact assessors and conservationists.”

Directions of Training

Morphology. Study of diagnostic characters of various groups of organisms. Sharma (2009) recommends “studying the morphological characters used in identification keys and practicing their recognition.”

Taxonomy. Understanding the principles of classification, nomenclature, and modern taxonomic concepts.

Molecular methods. Mastering methods of DNA extraction, PCR, sequencing, and bioinformatic analysis. Pandey (2021) emphasizes that “DNA barcoding is a new technique providing rapid and accurate species identification.”

Working with databases. Mastering searches in GenBank, BOLD, GBIF, Tropicos, and the use of bioinformatic tools.

Critical thinking. Developing the ability to critically evaluate identification results, including the results of AI tools.

Forms of Training

Courses and workshops. Participation in specialized courses on taxonomy, diagnostics, and molecular methods.

Internships. Work in herbaria, museums, and diagnostic laboratories.

Self-education. Study of literature, work with identification keys, practice of identification.

Participation in communities. Exchange of experience with colleagues, participation in conferences, publication of results.

Watson et al. (2015) note that “collaborative efforts allow taxonomists to appreciate the needs and demands of non-taxonomists,” and that “training and project development was also provided to partner institutions.”

Resources for Training

Literature. Identification keys, floras, monographs, textbooks.

Online resources. Watson et al. (2015) list: “The Internet Directory for Botany (IDB), The International Plant Names Index, JSTOR: Global Plants, Biodiversity Heritage Library, The USDA Plants Database, and many others.”

Databases. GenBank, BOLD, GBIF, Tropicos, CABI, EPPO.

Mobile applications. PlantNet, iNaturalist, Flora Incognita — as auxiliary tools for training.

Collections. Herbaria, museums, living collections of botanical gardens.

Daily Practice

Field observations. When detecting an organism:

1. Photograph it (several angles, with scale)

2. Record GPS coordinates, date, habitat

3. Collect a specimen (if possible and necessary)

4. Make field notes

Preliminary identification:

1. Use mobile applications to generate hypotheses

2. Test hypotheses using identification keys

3. Check ecological and geographic context

Verification:

1. Compare with the reference collection

2. Consult a specialist (if necessary)

3. Conduct molecular analysis (if necessary)

Documentation:

1. Record the identification method

2. Indicate the level of confidence

3. Preserve the specimen or photographs

4. Enter data into the database

Organization of Work

Creating a reference collection. Gradual accumulation of verified specimens in the course of work.

Maintaining a database. Systematic documentation of all identifications.

Cooperation. Establishing contacts with specialists and laboratories.

Training. Continuous advanced qualification.

Typical Errors and How to Avoid Them

Error 1: Relying only on AI. Solution: use AI as an additional tool, always verify.

Error 2: Ignoring ecological context. Solution: always check correspondence of habitat, host, and phenology.

Error 3: Not documenting. Solution: keep records, preserve specimens, indicate the method.

Error 4: Not admitting uncertainty. Solution: honestly indicate the level of confidence; when in doubt, consult.

Error 5: Using outdated names. Solution: check nomenclature in current databases.