Auto-ACMG Implementation#

Auto-ACMG operates through a structured process to determine the most appropriate predictor based on the type and context of the genetic variant being analyzed. This document describes the general workflow and the mechanisms behind the decisions that auto-acmg makes to predict the classification of genetic variants.

Workflow Overview#

  1. Variant Resolution: The first step involves determining the type of genetic variant. If the variant is recognized as a sequence variant (seqvar), auto-acmg employs the SeqVar default predictor. For structural variants (strucvar), the StrucVar default predictor is utilized, which primarily evaluates the PVS1 criteria.

  2. Gene-Specific Evaluation: For seqvars, auto-acmg checks if the variant is associated with any gene that has a specific Variant Curation Expert Panel (VCEP) implementation. If a specific VCEP is applicable, auto-acmg will switch to the corresponding VCEP predictor for that gene to leverage expert-specified evaluation criteria and thresholds.

  3. Structural Variants: For structural variants, the default predictor primarily evaluates the PVS1 criteria. You can find the implementation of the default StrucVar predictor in Implementation of different ACMG criteria for structural variants.

VCEP-Specific Overrides#

Each VCEP-specific predictor may override certain logic or thresholds of the Default Predictor:

  • Threshold Adjustments: Some VCEP implementations may just adjust thresholds for criteria like

    PP3 or BP4 and don’t override the prediction logic.

  • Prediction Logic Changes: Other VCEPs may override the entire logic for certain criteria. This

    can include changing how predictions are made based on the variant’s effects on protein function, splicing, or other molecular mechanisms.

Note

For the RYR1 gene, there are two VCEPs: Congenital Myopathies VCEP and Malignant Hyperthermia Susceptibility. In AutoACMG, we only consider the Malignant Hyperthermia Susceptibility VCEP for RYR1 and avoid the Congenital Myopathies VCEP. For more details on the VCEP implementations, please refer to the source code in the src/vcep directory of our GitHub repository.

Note

For von Willebrand Disease, there are two different rulesets. We implement the combination of both rulesets. For more details on the VCEP implementations, please refer to the source code in the src/vcep directory of our GitHub repository.

For a detailed overview of how specific VCEP predictors modify or extend the default behavior, you can review the source code on the GitHub repository under src/vcep and for the details refer to the ClinGen VCEP Criteria Specification Registry.