2.3. Mutational Data Analysis

Mutation Annotation Format (MAF) files used to store somatic variants detected were summarized, analyzed, annotated, and visualized using the maftools Bioconductor package [10]. Only variants assumed to have high or moderate (disruptive) impact in the protein, probably causing protein truncation, loss of function or triggering nonsense mediated decay were included in the analysis of most frequently mutated genes. For the calculation of tumor mutational load we considered both high/moderate impact mutation and all somatic mutations.

The DeconstructSigs package [11] was used to perform the mutational signature analysis. This tool evaluates the contribution of 30 signatures reported in COSMIC (https://cancer.sanger.ac.uk/cosmic/signatures) [12] to the mutational profile of each sample. Mutational signatures were calculated considering all somatic mutations in a given sample. The obtained signature scores were then analyzed in association with sensitivity class using Wilcoxon rank-sum test. Samples were grouped according to the top-5 most contributing mutational signatures using unsupervised hierarchical clustering performed with Euclidean distance and Ward linkage.

To identify mutations associated to sensitivity class we used the clinicalEnrichment function of maftools package [10] that performs Fisher’s exact tests to identify mutated genes associated with the class of interest. Analysis at the level of oncogenic pathways described in Sanchez-Vega et al. [13] was performed using the OncogenicPathways function of maftools. For each sample we classified each pathway as mutated if at least one of its genes carried a mutation. We then associated mutated pathways to sensitivity class using Fisher’s exact test. The same analysis was repeated using the “Hallmark” gene sets from MSigDB.