AnnotateMissense: a genome-wide annotation and benchmarking framework for missense pathogenicity prediction

Abstract

MotivationMissense variant interpretation remains challenging because pathogenicity depends on heterogeneous evidence, including population frequency, evolutionary conservation, transcript context, amino acid substitution severity, prior pathogenicity predictors and protein-language-model-derived features. Although these resources are individually useful, there remains a need for reproducible workflows that integrate them at genome-wide scale, benchmark their contribution and provide accessible outputs for downstream research. ResultsWe present AnnotateMissense, a scalable annotation, benchmarking and genome-wide prediction framework for missense variant interpretation. AnnotateMissense integrates chromosome-wise hg38 missense variants derived from dbNSFP v5.1 with ANNOVAR-based gene-, region- and filter-based annotations, dbNSFP transcript and protein descriptors, AlphaMissense scores, ESM-derived features, conservation metrics, population-frequency variables, established pathogenicity predictors and engineered amino acid/codon-context features. Using 132,714 ClinVar-labelled missense variants, we benchmarked machine-learning and deep-learning models under controlled feature configurations. The full 303-feature benchmark set achieved the strongest performance with XGBoost, reaching mean Matthews correlation coefficient (MCC) = 0.9411 and ROC-AUC = 0.9950 across stratified five-fold cross-validation. Restricted naive and location-oriented feature sets achieved substantially lower best MCC values of 0.4989 and 0.5113, respectively. Circularity-controlled ablations showed that removing prior-predictor, population-frequency and clinically overlapping evidence reduced performance, whereas excluding AlphaMissense and ESM-derived features alone had minimal effect. Temporal ClinVar validation on newly observed pathogenic/benign variants achieved MCC = 0.7613, accuracy = 0.8798 and F1-score = 0.8750. The final genome-wide model was applied to 90,643,830 hg38 missense variants to generate AnnotateMissense pathogenicity scores and binary prediction labels. Availability and implementationSource code, workflow scripts and command files are available at https://github.com/MuhammadMuneeb007/CAGI7_Annotate_All_Missense. Genome-wide prediction outputs and the compressed DuckDB database are available at https://doi.org/10.5281/zenodo.19981867. Supplementary informationSupplementary data are available online.