The Animal Variant Classification Guidelines v2: An Update With New Criteria and Improved Clarifications

ABSTRACT

The Animal Variant Classification Guidelines (AVCG) were developed to standardize and objectify the classification of putative disease‐causing variants. These guidelines are sufficiently reproducible and are used to classify previously published and new disease‐causing variants across species. Here, the guidelines are updated (AVCG.v2), based on a three‐phase decision process. Overall, four new criteria and seven clarifying comments were added. The number of criteria has increased from 23 to 27, with three new criteria supporting pathogenicity and one new criterion supporting benign classification. Pharmacogenomic variants were determined to fall within the scope of the guidelines. These updated guidelines are being used by the Variant Pathogenicity Working Group (VPWG), part of the Animal Genetic Testing Standardization standing committee, which is a committee of elected members of the International Society for Animal Genetics (ISAG). Under the auspices of ISAG, the VPWG retrospectively classifies published putative disease‐causing variants. The pathogenicity label for a variant will be presented in the variant tables of Online Mendelian Inheritance in Animals (OMIA; https://omia.org/). The AVCGv.2 criteria and recommendations were developed by the expertise of the animal genetics community and the ISAG Executive Committee through the Animal Genetics Testing Standardization Committee endorses and strongly encourages their use to evaluate the evidence supporting pathogenicity of putative disease‐causing variants.

The Animal Variant Classification Guidelines (AVCG, Boeykens et al. ²⁰²⁴) are now frequently used to classify putative disease‐causing variants (Abitbol et al. 2025; Boeykens et al. ²⁰²⁵; Eager, Jolly, et al. ²⁰²⁵; Eager, Willet, et al. ²⁰²⁵; Floriot et al. ²⁰²⁶; Jacinto, Letko, et al. ²⁰²⁵; Jacinto, Leuenberger, et al. ²⁰²⁵; Kaczmarska et al. ²⁰²⁵; Stee et al. ²⁰²⁵). This classification process entails evaluating a set of criteria (e.g., with in silico tools, assessing genotype–phenotype segregation patterns), followed by applying a decision tree, which leads to an assigned pathogenicity label (pathogenic [P], likely pathogenic [LP], variant of uncertain significance, likely benign, or benign). Each label is associated with a recommendation regarding the use of that variant in clinical‐decision making, for screening purposes and breeding (Boeykens et al. 2024; Casselman et al. ²⁰²⁶).

The guidelines can be used prospectively (i.e., during the identification and publication of new putative disease‐causing variants) and retrospectively (i.e., when evaluating the evidence on previously published putative disease‐causing variants), thereby objectifying the evaluation of putative disease‐causing variants. A systematic ongoing evaluation of published putative disease‐causing variants was established by the Variant Pathogenicity Working Group (VPWG), under the auspices of the Animal Genetic Testing Standardization Standing Committee of the International Society for Animal Genetics (ISAG), in close collaboration with the Online Mendelian Inheritance in Animals (OMIA) database (Nicholas et al. 2025; Tammen et al. ²⁰²⁴). So far, pathogenicity labels have been assigned for 149 variants across three species (dog, cat, horse) and these have been included in the OMIA variant tables (Casselman et al. 2026).

Building on the experience gained by using these guidelines, this study presents the first update of the AVCG (Boeykens et al. 2024). The VPWG, a multidisciplinary group of veterinarians and geneticists from universities and commercial laboratories, worked in three phases to improve the first version of the AVCG (AVCG.v1), published in December 2024 (Boeykens et al. 2024). As during the development of AVCG.v1, this process combined (anonymous) individual input, and a focus group discussion (Appendix S1). Proposals were incorporated when a two‐third majority was reached.

Sixteen changes were suggested, of which one related to the scope, seven to the criteria (five new criteria, two suggestions for existing criteria), seven to the clarifications, and one to the decision tree (Appendix S2). Ultimately, 11 changes achieved a two‐thirds majority and, consequently, were incorporated, comprising four changes to the criteria (i.e., four new criteria added, namely PM5, PM6, PP5 and BS4) and seven to the clarifications. An overview of the updated criteria (AVCG.v2) is provided in Table 1, whereas the updated clarifications are provided in Appendix S3. AVCG.v2 contains 27 criteria compared to the 23 in AVCG.v1. The decision tree can be found in Table 2.

TABLE 1.

The criteria in the updated version of the Animal Variant Classification Guidelines (AVCG.v2).

Name	Criterion
PVS1	Null variant (nonsense, frameshift, canonical ±1 or 2 splice‐sites, initiation codon, single or multi‐exon deletion) in a gene where LOF is a known mechanism of disease in the same or another species, if functionality of the gene is expected to be similar across species
PS1	Same amino acid change as a previously established pathogenic variant regardless of nucleotide change
PS2	De novo in a patient with the disease and unaffected parental samples tested negative
PS3	Well‐established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product
PS4	The prevalence of the variant in affected individuals is significantly increased compared with the prevalence in controls
PS4_moderate	See clarification section ( Appendix S3 )
PS5	Cosegregation with disease in multiple affected family members in a gene definitively known to cause the disease
PS5_moderate	See clarification section ( Appendix S3 )
PM1	Located in a mutational hot‐spot and/or critical and well‐established functional domain (e.g., active site of an enzyme) without benign variation across breeds and/or species
PM2	Novel missense change at an amino acid residue where a different missense change has been determined to be pathogenic in other individuals
PM3	For recessive disorders, detected in trans with a pathogenic variant
PM4	Protein length changes as a result of in‐frame deletions/insertions in a non‐repetitive region or stop‐loss variants
PM5	The causal variant has been reliably mapped to an interval ≤ 3 Mb. Examples include statistically significant GWAS signals or LOD scores > 3 in linkage analyses. GWAS and linkage may be further refined by haplotype analyses (e.g., autozygosity mapping), but shared haplotypes without significant association to the disease are not sufficient to apply this criterion
PM6	Genotype–phenotype association is confirmed in multiple breeds, in addition to the original population in which the association was identified
PM6_supporting	See clarification section ( Appendix S3 )
PP1	Cross‐species alignment shows the variant is conserved and other information across species (e.g., ClinVar data) states the variant is pathogenic
PP2	Missense variant in a gene that has a low rate of benign missense variation and in which missense variants are a common mechanism of disease
PP3	All computational evidence supports a deleterious effect on the gene or gene product (conservation, evolutionary, splicing impact, etc.)
PP4	Patient's phenotype or family history is highly specific for a disease with a single genetic etiology
PP5	Disease‐associated genotype is absent or rare in appropriate controls and/or population databases
BS1	Lack of segregation in affected members of a family
BS2	Observed in a healthy adult individual for a recessive (homozygous), dominant (heterozygous), or X‐linked (hemizygous) disorder, with full penetrance expected at an early age
BS3	Well‐established in vitro or in vivo functional studies show no damaging effect on protein function or splicing
BS4	When the sample size is sufficient, the prevalence of the disease‐causing variant in affected individuals is similar (e.g., odds ratio confidence interval includes 1) to the prevalence in controls
BP1	Cross‐species alignment shows the variant is not conserved and other information across species (e.g., ClinVar data) states the variant is benign
BP2	Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder or observed in cis with a pathogenic variant in any inheritance pattern
BP3	In‐frame deletions/insertions in a repetitive region without a known function
BP4	All computational evidence supports a benign effect on the gene or gene product (conservation, evolutionary, splicing impact, etc.)
BP5	Variant found in a case with an alternate molecular basis for disease
BP6	A synonymous (silent) variant for which splicing prediction algorithms predict no impact to the splice consensus sequence nor the creation of a new splice site AND the nucleotide is not highly conserved

Note: Changes in version 2 relative to version 1 are underlined. The name of each criterion designates whether it supports pathogenic or benign classification of a variant (P or B, respectively), followed by the weight of the support (VS = very strong; S = strong; M = moderate; P = supportive, respectively) and a number that represents the criterion. Explanations per criterion are provided in Appendix S3.

Abbreviations: GWAS, genome‐wide association study; LOD, logarithm of the odds; LOF, loss‐of‐function.

TABLE 2.

The decision tree used to assign a genetic variant pathogenicity label.

Step 1: Always check branch A and branch B
Branch A: Pathogenic variant
Pathogenic (P)	Very strong (PVS1) AND ≥ 1 strong (PS1–PS5) OR ≥ 2 moderate (PM1–PM6) OR 1 moderate (PM1–PM6) AND 1 supporting (PP1–PP5) OR ≥ 2 supporting (PP1–PP5) ≥ 2 strong (PS1–PS5) 1 strong (PS1–PS5) AND ≥ 3 moderate (PM1–PM6) OR 2 moderate (PM1–PM6) AND ≥ 2 supporting (PP1–PP5) OR 1 moderate (PM1–PM6) AND 4 supporting (PP1–PP5)
Likely pathogenic (LP)	Very strong (PVS1) AND 1 moderate (PM1–PM6) 1 strong (PS1–PS5) AND 1–2 moderate (PM1–PM6) 1 strong (PS1–PS5) AND ≥ 2 supporting (PP1–PP5) 3 moderate (PM1–PM6) 2 moderate (PM1–PM6) AND ≥ 2 supporting (PP1–PP5) 1 moderate (PM1–PM6) AND 4 supporting (PP1–PP5)
Branch B: Benign variant
Benign (B)	≥ 2 strong (BS1–BS4)
Likely benign (LB)	1 strong (BS1–BS4) AND 1 supporting (BP1–BP6) ≥ 2 supporting (BP1–BP6)
Step 2: Variants without a label or with a label from the pathogenic (P/LP) and benign (B/LB) branch
Variant of uncertain significance (VUS)	Not enough criteria met to assign label Labels from two conflicting branches assigned

Note: To assign a pathogenicity classification label, all criteria in Table 1 have to be evaluated. The number of criteria fulfilled in each weight category are counted. Once all criteria have been evaluated and a final count per weight category has been obtained, a variant pathogenicity classification label is assigned. This is a stepwise process in which first both branch A and branch B have to be considered. If no classification is assigned or if a classification from branch A and B is assigned, step two also has to be considered.

Some decisions warrant additional clarifications. The first one was linked to the scope. It was questioned whether pharmacogenomic variants should be classified because they only cause disease after a trigger, that is, medication. They are now considered to fall in the scope in AVCG.v2, on the grounds of their potential serious impact on animal health and the observation that environmental stimuli might also be needed for other variants that were classified (e.g., for variants causing coagulopathies to exert their effect, quite often the patient needs to suffer a trauma). The proposal to explicitly rule them out in the scope text, was thus not accepted.

Of the four new criteria (PM5, PM6, PP5, and BS4; Table 1), three support pathogenicity. One of these, PP5 (Table 1), reintroduces the use of allele frequencies, which was initially removed in AVCG.v1 (Boeykens et al. 2024; Richards et al. ²⁰¹⁵). One new benign strong criterion is added (BS4, Table 1), stating that when the prevalence of the disease‐causing variant in affected individuals is similar to the prevalence in controls, this is evidence against pathogenicity.

New in AVCG.v2 is that three criteria (PS4, PS5, PM6) can be assigned a lower weight of evidence, that is, PS4 and PS5 can now also be graded as moderate; and PM6 can now also be graded as supporting (Table 1) (Richards et al. 2015). These new criteria retain their original name, followed by an underscore and the assigned weight, that is, PS4_moderate, PS5_moderate and PM6_supporting.

A suggested change that was not incorporated was to make rules to exclude new combinations of criteria, aside from the ones that were mutually exclusive. The reason for the proposal was the observation that some criteria are highly correlated, which might potentially make it too easy to reach the P/LP threshold and may encourage reuse of the same data/arguments for different criteria. Although this change was not accepted, a list that explicitly defines which criteria already were mutually exclusive was added (Appendix S4). Furthermore, the use of the same data for different criteria should be avoided.

Several challenges still lie ahead. Currently, AVCG.v2 has been developed with broad applicability in mind. Although phenotype‐specific guidelines have not yet been discussed, this might be necessary in the future. In human genetics, individual gene‐ and/or phenotype‐specific guidelines have been established. Examples include guidelines for pathogenic variants in BMPR2 associated with pulmonary arterial hypertension, as well as for variants in a suite of genes associated with severe combined immunodeficiency (Eichstaedt et al. 2025; Jacovas et al. ²⁰²⁶). In addition, non‐coding variants remain difficult to classify with the current guidelines. Furthermore, the VPWG has concluded that in silico tools are currently only sufficiently accurate to assess two types of variants (missense and splice site variants [Appendix S3, PP3/BP4; Boeykens et al. ²⁰²⁴]). As new tools are published, updates to the guidelines are expected.

To conclude, an updated version of the AVCG (i.e., AVCG.v2) is presented, with modifications at the level of the criteria, as well as the clarifications. These guidelines will improve the classification process further, whereas keeping the aforementioned limitations in mind. The manual used by the VPWG is available in Appendix S5. The AVCG.v2 criteria and recommendations were developed by the expertise of the animal genetics community and the ISAG Executive Committee endorses and strongly encourages their use to evaluate the evidence supporting pathogenicity of putative disease‐causing variants. Consequentially, only variants with LP or P labels, should be used in clinical‐decision making, for screening purposes and in breeding decisions.

Author Contributions

Lore Desmet: investigation, project administration, resources, writing – original draft. Frank W. Nicholas: investigation, formal analysis, writing – review and editing. Pascale Smets: investigation, writing – review and editing, formal analysis. Bart J. G. Broeckx: funding acquisition, visualization, supervision, writing – review and editing, software, formal analysis, conceptualization, data curation. Carlotta Ferrari: investigation, formal analysis, writing – review and editing. Heidi Anderson: investigation, formal analysis, writing – review and editing. Caroline Dufaure de Citres: investigation, formal analysis, writing – review and editing. Nüket Bilgen: formal analysis, writing – review and editing. Marie Abitbol: investigation, formal analysis, writing – review and editing. Jonas Donner: formal analysis, writing – review and editing. Danika Bannasch: investigation, formal analysis, writing – review and editing. Jerold Bell: investigation, formal analysis, writing – review and editing. Steven M. Harrison: formal analysis, writing – review and editing. Jens Häggström: investigation, formal analysis, writing – review and editing. Ingrid Ljungvall: investigation, formal analysis, writing – review and editing. Jessica J. Hayward: investigation, formal analysis, writing – review and editing. Jason T. Huff: investigation, formal analysis, writing – review and editing. Guillermo Giovambattista: formal analysis, writing – review and editing. Carrie J. Finno: investigation, formal analysis, writing – review and editing. Leslie A. Lyons: investigation, formal analysis, writing – review and editing. Maria Longeri: investigation, formal analysis, writing – review and editing. Åsa Ohlsson: investigation, formal analysis, writing – review and editing. Lucie Chevallier: formal analysis, writing – review and editing. Samantha Van Buren: formal analysis, writing – review and editing. Cathryn Mellersh: investigation, formal analysis, writing – review and editing. Frank G. van Steenbeek: investigation, formal analysis, writing – review and editing. Jessica L. Petersen: formal analysis, writing – review and editing. Hubert Bauer: investigation, formal analysis, writing – review and editing. Maria G. Strillacci: investigation, formal analysis, writing – review and editing. Marcela Martinez: investigation, formal analysis, writing – review and editing. Imke Tammen: investigation, formal analysis, writing – review and editing, data curation. Tosso Leeb: investigation, writing – review and editing.

Funding

This work was supported by Royal Canin and Ronald Bruce Anstee Bequest.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

For a part of the study period, H.A. was an employee of Wisdom Panel Petcare Science & Diagnostics, a company that offers canine and feline DNA testing as a commercial service. H.B. is an employee of Laboklin, a DNA testing and genetic analysis company for dogs, cats, horses, and wildlife. N.B. is an employee of GeneControl, a DNA testing and genetic analysis company for dogs, cats, horses, and wildlife. C.D.C. is an employee of Antagene, a DNA testing and genetic analysis company for dogs, cats, horses, and wildlife. J.D. is the owner of GenomeTails, a genetic counselling company. S.M.H. is an employee of Ambry Genetics, a DNA testing and genetic analysis company for humans. J.T.H. is an employee of Wisdom Panel Mars Petcare Science & Diagnostics, a company that offers canine and feline DNA testing as a commercial service. S.V.B. is an employee of Orivet, a DNA testing and genetic analysis company for dogs and cats. A portion of the revenue generated by UC Davis Veterinary Genetics Lab (UCD‐VGL) for tests discovered in the Finno Laboratory (MYHM and EJSCA) provides funding for additional research conducted in the Finno Laboratory for 10 years from the test launch date. TL's research lab receives revenues from a patent on genetic testing for hereditary nasal parakeratosis in Labrador Retrievers and occasional income from services provided to commercial genetic testing laboratories. This does not influence any of the results generated in the current study.

Supporting information

Appendix S1: Overview of the three‐phase decision process used to adapt the Animal Variant Classification Guidelines (AVCG).

Appendix S2: Overview of all the changes that were discussed to obtain the updated version of the Animal Variant Classification Guidelines (AVCG).

Appendix S3: Clarifications of the criteria of the updated Animal Variant Classification Guidelines (AVCG.v2). Changes to the clarifications relative to version 1 of the AVCG are underlined.

Appendix S4: List of the criteria that are mutually exclusive in the updated Animal Variant Classification Guidelines (AVCG.v2).

Appendix S5: The variant classification manual.

Data Availability Statement

The data that supports the findings of this study are available in the main text and Supporting Information of this article.