Somatic mutation detection in plants: a critical evaluation through simulations and reanalyses
Abstract
Mutation, the source of genetic diversity, is the raw material of evolution; however, the mutation process remains understudied, especially in plants. Using both a simulation and reanalysis framework, we set out to test the performance of two types of variant callers, generic ones and those developed for cancer research, to detect de novo somatic mutations. In an in silico experiment, we generated Illumina-like sequence reads spiked with simulated mutations at different allele frequencies to compare the performance of seven commonly-used variant callers to recall them. More empirically, we then reanalyzed two of the largest datasets available for plants, both developed for identifying within-individual variation in long-lived pedunculate oaks. Even in plants, variant callers developed for cancer research outperform generic callers regarding mutation recall and precision, especially at low allele frequency. Such variants at low allele frequency are typically expected for within-individual de novo plant mutations. Reanalysis of published oak data with the best-performing caller based on our simulations identified up to 7x more somatic mutations than initially reported. Our results advocate the use of cancer research callers to boost de novo mutation research in plants, and to reconcile empirical reports with theoretical expectations.