Supplementary tables.

Table A in S1 Table: The cosine similarities between the de novo signatures and the true in vitro signatures. De novo signatures are discovered by SUITOR, SUITOR, SigProfilerExtractor, SignatureAnalyzer and signeR. Cosine similarity is used to compare them with true in vitro signatures: the background signature and signature induced by MSH6 gene knockout. Cosine similarity ranges from 0 to 1, with a cosine of 1 indicating a perfect match. Table B in S1 Table: The cosine similarities between the signatures discovered by SUITOR, SigProfilerExtractor, SignatureAnalyzer, signeR and the true mutagen-induced signatures. For each de novo signature, the largest cosine similarities are highlighted in red, indicating the strongest similarity to the corresponding in vitro signatures. ENU: N-ethyl-N-nitrosourea; DES: diethyl sulfate; 1,2-DMH+S9: 1,2-dimethylhydrazine mixed with S9 rodent liver-derived metabolic enzyme; TMZ: Temozolomide; SSR: simulated solar radiation; 6-NC: 6-Nitrochrysene; AAI: aristolochic acid I. Table C in S1 Table: The percentage of nonzeros, mean and standard deviation (SD) of signature contributions in in silico simulation studies of nine signatures. Table D in S1 Table: The selected number of signatures out of 20 in in silico simulations of nine signatures for SUITOR, SigProfilerExtractor, SignatureAnalyzer, signeR. The column with #(CS > 0.8) indicates the number of signatures based on the threshold 0.8, while the column with #(CS>0.9) indicates the number of signatures based on the threshold 0.9. Table E in S1 Table: The selected number of signatures out of 20 in in silico simulations with sequencing errors for SUITOR, SigProfilerExtractor, SignatureAnalyzer, signeR. The column with #(CS > 0.8) indicates the number of signatures based on the threshold 0.8, while the column with #(CS>0.9) indicates the number of signatures based on the threshold 0.9. Sequencing errors per each tumor were generated from a uniform distribution within the range [0, a×b], where the letter a denotes ErrorLevel (0 means no sequencing errors) and b is the average mutation count of each tumor. The frequencies of signature SBS9 are highlighted in red because the frequency of signature SBS9 to be detected across 20 replicates would be reduced for all methods if the cosine similarity threshold is increased to 0.9. Table F in S1 Table: The selected number of signatures out of 20 in in silico simulations with sequencing errors for SparseSignatures and CV2K. Sequencing errors per each tumor were generated from a uniform distribution within the range [0, a×b], where the a denotes ErrorLevel (0 means no sequencing errors) and b is the average mutation count of each tumor. Table G in S1 Table: The number of signatures identified by SparseSignatures and CV2K for eight cancer types of PCAWG studies, entire PCAWG data and Sanger Breast cancer study. The column with CS* > 0.8 indicates the number of signatures whose largest cosine similarity is greater than 0.8 for SparseSignatures method. Table H in S1 Table: The cosine similarities between de novo signatures discovered by SparseSignatures in in vivo studies. We annotated them with COSIMC signature having the largest cosine similarity. COSMIC signatures and cosine similarities are highlighted in red if it is larger than 0.8. Duplicated signatures with repsect to the highest cosine similarity are marked with an asterisk. Table I in S1 Table: The cosine similarity between de novo signatures discovered by SUITOR in PCAWG breast cancer study and de novo signatures discovered by SUITOR in Sanger breast cancer study. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table J in S1 Table: The cosine similarity between de novo signatures discovered by signeR in PCAWG breast cancer study and de novo signatures discovered by signeR in Sanger breast cancer study. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table K in S1 Table: The cosine similarity between de novo signatures discovered by SigProfilerExtractor in PCAWG breast cancer study and de novo signatures discovered by SigProfilerExtractor in Sanger breast cancer study. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table L in S1 Table: The cosine similarity between de novo signatures discovered by SignatureAnalyzer in PCAWG breast cancer study and de novo signatures discovered by SignatureAnalyzer in Sanger breast cancer study. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table M in S1 Table: The cosine similarity between de novo signatures discovered by SUITOR in PCAWG breast cancer study and in Sanger breast cancer study with COSMIC signature profiles. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table N in S1 Table: The cosine similarity between de novo signatures discovered by SigProfilerExtractor in PCAWG breast cancer study and in Sanger breast cancer study with COSMIC signature profiles. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table O in S1 Table: The cosine similarity between de novo signatures discovered by SignatureAnalyzer in PCAWG breast cancer study and in Sanger breast cancer study with COSMIC signature profiles. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table P in S1 Table: The cosine similarity between de novo signatures discovered by signeR in PCAWG breast cancer study and in Sanger breast cancer study with COSMIC signature profiles. For each column, the largest cosine similarity is highlighted in red if it is larger than 0.8 or in blue otherwise, corresponding to the COSMIC signature most similar to a given PCAWG or Sanger signature. Table Q in S1 Table: The mean and standard deviation (SD) of signature contributions by signature subtype G3 and G4. Table R in S1 Table: The clinical and pathological factors associated with the subgroups G3 and G4. a) For continuous factors; b) for discrete factors. Table S in S1 Table: The cosine similarities between de novo INDEL signatures in ovarian adenocarcinoma by SUITOR and COSIMC INDEL signatures. Table T in S1 Table: The cosine similarity between de novo signatures discovered by SUITOR and the signatures reported in the original C. elegans study (Volkova et al. 2020). COSMIC signatures and cosine similarities are highlighted in red if it is larger than 0.8. Table U in S1 Table: The cosine similarity between de novo signatures discovered by SUITOR in PCAWG colorectal, esophageal, and head SCC cancer studies. COSMIC signatures and cosine similarities are highlighted in red if it is larger than 0.8. The numbers in parentheses are the number of tumors for each study.
(XLSX)