Fig. 5
From: Bayesian multi-study non-negative matrix factorization for mutational signatures
A (Top) Signature indicator matrix estimated using the recovery-discovery model with covariates in the colorectal cancer application. Signatures in the discovery component are labeled beginning with “D.” (Bottom) As above, but showing the proportion of samples in each study with more than 5% of total exposures attributed to each signature. B Posterior inclusion probabilities for each covariate and signature in the ultra-hypermutated group