Additional file 1 of Mutation-Attention (MuAt): deep representation learning of somatic mutations for tumour typing and subtyping

Additional file 1: Table S1. Dataset Information for PCAWG, TCGA, ICGC, and common tumour subtypes in PCAWG and TCGA datasets. Table S2. Precision, recall, and f1 scores in the PCAWG and TCGA datasets. Table S3. Classification accuracy of the different models benchmarked. Table S4. Coefficients with 95% confidence intervals, p-values and FDR-adjusted q-values of a PCA model associating driver mutation statuses with MuAt features in the PCAWG dataset tumours. Table S5. Coefficients with 95% confidence intervals, p-values and FDR-adjusted q-values of a negative binomial model associating driver mutation statuses with mutation counts in the PCAWG dataset tumours. Table S6. Coefficients with 95% confidence intervals, p-values and FDR-adjusted q-values of a least-squares linear regression model associating MuAt tumour-level features with the fraction of mutated microsatellitesin the PCAWG dataset tumours. Table S7. Coefficients with 95% confidence intervals, p-values and FDR-adjusted q-values of a negative binomial model associating MuAt tumour-level features with mutation counts in the PCAWG dataset tumours. Table S8. Pancreatic neuroendocrine tumours in the PCAWG dataset with germline MUTYH driver mutations. Table S9. Linear association of MuAt factors with mutation counts. Table S10. Linear association of MuAt factors with mutational signatures. Table S11. Attention values for mutation type pairs, averaged over tumour types.