The PM1 rule In the 2015 ACMG/AMP sequence variant interpretation guidelines, is related to missense variants located in a mutational hot spot and/or critical and well-established functional domain. These regions usually lack benign (B) non-truncating (missense, inframe indels) variants while enriched with pathogenic (P) non-truncating variants. While this functional data evidence type could be very helpful in the classification of difficult variants such as missense or inframe variants, there are current analytical challenges specific to hotspot regions, e.g., no existing hotspot region databases, or gold standard.

Franklin uses the Genoox AI Engine in order to detect such hotspot regions and apply the PM1 rule accordingly.

For each exon or domain in a gene, a sliding window first extracts candidate regions between each pair of benign (B) variants in order for it to be clear of B variants. Candidate regions without pathogenic (P) variants within them are ignored. Within each candidate region, the AI model further detects the inner borders of P non-truncating variants contained within, and the number of P variants. Based on the density and number of P variants, the model evaluates each region for the presence of a hotspot and then assigns a weight to the PM1 rule as ‘supporting’(PP), ‘moderate’ (PM), or ‘strong’ (PS).

In the absence of a gold standard for hotspot regions, we benchmarked the Variant Classification Engine (VCE) ability to detect hotspot regions against the different gene/disease-specific expert panels’ (EP’s) recommendations for PM1 and their published curated variant datasets.

More details can be found in the poster “Using Artificial Intelligence for Detection of Hotspot Regions: ACMG/AMP PM1 Criterion” we presented in the 2019 ACMG conference.