2.2.2. Mutation-Level Features

Features on a mutation-level were computed using the web interface of Variant Effect Predictor (VEP) based on Ensembl genome database (release 100, April 2020), which annotates genomic variants based on their consequence [26]. Sorting Intolerant From Tolerant (SIFT) and Polymorphism Phenotyping (PolyPhen) scores were chosen, as they show the potential impact of amino acid substitutions on the protein function [27,28], while Consensus Deleteriousness (Condel) scores indicate the probability of a single base mutation of being deleterious [29]. Thus, a SIFT score of 0 represents the most deleterious variant while a PolyPhen or a Condel score of 1 shows the variant with the highest deleterious potential. Additional rank scores were extracted from the database for Nonsynonymous SNPs’ Functional Predictions (dbNSFP), which provides pathogenicity scores from multiple popular algorithms [30]. The highest rank score for each mutation was chosen and then the average value between all scores was calculated based on the rationale that the highest rank score will show the potentially most damaging effect, while the arithmetic mean will reflect the overall pathogenic consequence. A full list of all rank scores used for the computation can be found in Supplementary Table S2: List of features used to compute the average rank score.