Correlation of Mutational Signatures in Cancer Genes with General Signatures

The occurrence of various mutation patterns, such as changes in the DNA sequence and the loss of some sequences, is called a “mutational signature,” and they represent the molecular fingerprints that exist for the type of mutation occurring in a specific gene. Our study elucidates the correlations of mutational signatures in frequently mutated cancer genes with general mutational signatures previously found for different cancers. We hypothesized that the top twenty most frequently mutated genes (MFMG) of a cancer type would have the highest correlation with the general signatures related to the cancer. The program for our research, SignaGen, was created using MATLAB to take in genomic sequence data and mutation data (consisting of the type, location, and frequency of each mutation) to calculate the mutational signatures of genes. The correlation values for the top twenty MFMG were organized into heatmaps for each cancer type observed. By looking at the heatmaps, we could see that the MFMG did have relatively higher correlation values with the general signatures that were related to the cancer type. However, there were also cases in which the MFMG had lower correlation values with a related signature than other associated signatures. For example, the MFMG of skin cancer had an average correlation of 7.76% with signature 17, while having an average correlation of 43.02% with signature 11. Not only was the average lower than the average correlation with the other related general signatures, but it was also lower than the average correlation with unrelated signatures. To investigate this inconsistency and verify the significance of these correlation values, we compared the correlation values of the MFMG to the correlation values of randomly selected genes of similar length. Even if the MFMG’s correlation with the related signatures is low, our hypothesis would still be supported if they had a higher correlation than the random genes. We took three of the twenty MFMG for each cancer and compared their correlation with the random genes of similar length for each cancer type and found that the MFMG had a higher correlation than the random genes for most cases.