Accurate detection of somatic single-nucleotide variants from bulk RNA-seq data using RNA-MosaicHunter

Abstract

Somatic variants are increasingly recognized as contributors to diverse non-cancer, developmental, and aging-related disorders. However, most tools for detecting somatic single-nucleotide variants (sSNVs) were designed for DNA sequencing and primarily tailored to cancer datasets, leaving a critical gap in harnessing the rich potential of RNA-seq for sSNV identification, particularly in non-cancer tissues with low mutation rates. Here, we introduce RNA-MosaicHunter, a novel bioinformatic tool for accurate sSNV detection from bulk RNA-seq. In two benchmarking datasets, it demonstrated high precision (94.7% in TCGA and 99.3% in a cell-line mixture) with sensitivities of 53.4% and 38.9%, respectively, in the default mode that maximizes precision. We then applied RNA-MosaicHunter to profile 827 RNA-seq samples in three tissue types from the Genotype Tissue Expression project (GTEx), where it outperformed previous methods in capturing mutational characteristics associated with normal aging. We further utilized RNA-MosaicHunter to analyze RNA-seq data from 382 Alzheimer's disease (AD) brain samples and 480 age-matched controls and revealed a significantly higher burden of sSNVs in AD cerebral cortex, suggesting the potential contribution of sSNVs to AD pathogenesis. RNA-MosaicHunter enables accurate profiling and characterization of sSNVs from RNA-seq data, advancing the understanding of the role of somatic variants across diverse tissues and diseases.