mSigHdp: hierarchical Dirichlet process mixture modeling for mutational signature discovery

Abstract

Mutational signatures are characteristic patterns of mutations caused by endogenous or exogenous mutational processes. These signatures can be discovered by analyzing mutations in a large set of samples – usually somatic mutations in tumor samples. Most approaches to mutational-signature discovery are based on non-negative matrix factorization. Alternatively, signatures can be inferred using hierarchical Dirichlet process (HDP) mixture models, an approach that has been relatively little explored. These models assign mutations to clusters and view each cluster of mutations as being generated from a particular mutational process. Here we describe mSigHdp, an improved approach to using HDP mixture models to discover mutational signatures. We benchmarked mSigHdp and several other programs on realistic synthetic data. For single-base mutations, mSigHdp and the widely used SigProfilerExtractor discovered signatures better than the other 2 programs and had different strengths: mSigHdp discovered more of the signatures present in the synthetic data, while SigProExtractor discovered fewer false positives. mSigHdp was better able to discover rare signatures, which may be an advantage, since most common signatures have probably already been discovered. For small insertion-and-deletion mutations, mSigHdp discovered signatures better than the other 3 programs. Thus, mSigHdp is an advance for discovering rare SBSs mutational signatures and for discovering small insertion-and-deletion mutational signatures.AvailabilitymSigHdp is available at GitHub public repositories https://github.com/steverozen/mSigHdp and https://github.com/steverozen/hdpx. Synthetic data, code for generating the synthetic data, code for running the mutational-signature discovery programs and analyzing their results, the main outputs of the programs, and code for generating the data figures in this paper are available at https://github.com/Rozen-Lab/Liu_et_al_Sup_Files.Supplementary informationSupplementary information is provided on the bioRxiv web site for this manuscript.