varCADD: large sets of standing genetic variation enable genome-wide pathogenicity prediction.

Abstract

Background: Machine learning and artificial intelligence are increasingly being applied to identify phenotypically causal genetic variation. These data-driven methods require comprehensive training sets to deliver reliable results. However, large unbiased datasets for variant prioritization and effect predictions are rare as most of the available databases do not represent a broad ensemble of variant effects and are often biased towards the protein-coding genome, or even towards few well-studied genes.

Methods: To overcome these issues, we propose several alternative training sets derived from subsets of human standing variation. Specifically, we use variants identified from whole-genome sequences of 71,156 individuals contained in gnomAD v3.0 and approximate the benign set with frequent standing variation and the deleterious set with rare or singleton variation. We apply the Combined Annotation Dependent Depletion framework (CADD) and train several alternative models using CADD v1.6.

Results: Using the NCBI ClinVar validation set, we demonstrate that the alternative models have state-of-the-art accuracy, globally on par with deleteriousness scores of CADD v1.6 and v1.7, but also outperforming them in certain genomic regions. Being larger than conventional training datasets, including the evolutionary-derived training dataset of about 30 million variants in CADD, standing variation datasets cover a broader range of genomic regions and rare instances of the applied annotations. For example, they cover more recent evolutionary changes common in gene regulatory regions, which are more challenging to assess with conventional tools.

Conclusions: Standing variation allows us to directly train state-of-the-art models for genome-wide variant prioritization or to augment evolutionary-derived variants in training. The proposed datasets have several advantages, like being substantially larger and potentially less biased. Datasets derived from standing variation represent natural allelic changes in the human genome and do not require extensive simulations and adaptations to annotations of evolutionary-derived sequence alterations used for CADD training. We provide datasets as well as trained models to the community for further development and application.