Fig. 1: Somatic mutation phenotypes in ~15,000 human tumors.
From: The impact of rare germline variants on human somatic mutation processes
a Somatic mutations were extracted from approximately 9300 whole-exome and 5500 whole-genome sequenced cancer genomes (left). 56 different somatic mutation features were estimated in each cancer genome, covering different types of mutations (right table). b Final set of somatic components was extracted by applying two methods to the input matrix (tumor samples as rows and somatic mutation features as columns): independent component analysis (ICA) and a variational autoencoder (VAE). 15 ICA-derived and 14 VAE-derived components (mutation phenotypes) were extracted. c Overview of extracted somatic mutation components (x-axis) and their Pearson correlation (color code) with the input somatic mutation features (y-axis). Gray strip at the bottom displays whether the component was extracted via ICA or VAE. Components were named based on the underlying mutational process or strongest correlating input feature(s). dMMR, deficient DNA mismatch repair, dHR, deficient homologous recombination. Data underlying panel c are provided as a Source Data file.
