Whole-Genome Analysis of De Novo Somatic Point Mutations Reveals Novel Mutational Biomarkers in Pancreatic Cancer

Abstract

It is now known that at least 10% of samples with pancreatic cancers (PC) contain a causative mutation in the known susceptibility genes, suggesting the importance of identifying cancer-associated genes that carry the causative mutations in high-risk individuals for early detection of PC. In this study, we develop a statistical pipeline using a new concept, called gene-motif, that utilizes both mutated genes and mutational processes to identify 4211 3-nucleotide PC-associated gene-motifs within 203 significantly mutated genes in PC. Using these gene-motifs as distinguishable features for pancreatic cancer subtyping results in identifying five PC subtypes with distinguishable phenotypes and genotypes. Our comprehensive biological characterization reveals that these PC subtypes are associated with different molecular mechanisms including unique cancer related signaling pathways, in which for most of the subtypes targeted treatment options are currently available. Some of the pathways we identified in all five PC subtypes, including cell cycle and the Axon guidance pathway are frequently seen and mutated in cancer. We also identified Protein kinase C, EGFR (epidermal growth factor receptor) signaling pathway and P53 signaling pathways as potential targets for treatment of the PC subtypes. Altogether, our results uncover the importance of considering both the mutation type and mutated genes in the identification of cancer subtypes and biomarkers.

Keywords: cancer subtype identification; genotype and phenotype characterization; pancreatic cancer; personalized medicine; somatic point mutations; therapeutic targets.

Figure 1

The workflow of pancreatic cancer subtype identification and clustering tree. In the top left, an overall view of the 3-mer motif and the gene-motif concept is illustrated. (a) At first, we construct features named gene-motifs based on the 3-mer motif and the gene that motif has occurred in. These features were constructed for all samples and in all of their protein-coding genes. In the top right, the feature selection process is illustrated. (b) We calculated the number of samples each gene-motif has occurred in, and based on their distributions, we found the most frequent (and hence significant) gene-motifs. We also found the most frequent mutated genes or significantly mutated genes to filter out those gene-motifs that have not occurred in significant genes. This leads to significant features for clustering. (c) The clustering process and tree is illustrated. After constructing a matrix of occurrence for each feature in each sample, (each cell indicates whether a feature has occurred in a sample or not) the Mclust algorithm was employed to cluster samples into subtypes. After two rounds of clustering, five main subtypes revealed themselves. (d) Finally, comprehensive genotype and phenotype characteristic study was performed to find differences and/or commonality in subtypes (bottom left). This includes gene association, mutational signature, deep mutational profile investigation, finding DEGs, survival analysis, etc.

Figure 2

The mutation rate in subtypes and their differences. Bar plots in green tiles exhibit mutation frequency in protein-coding genes and significant features. Yellow tiles include bar plots of differences of mutation rate for significant features and blue tiles include differences in mutation rate in protein-coding genes. For example, the bar plot in the tile which is in PCS2 column and PCS2 row represents the difference of mutation rate in protein-coding genes in PCS2 and PCS1. The color of bars in differential bar plots represents the subtype with the higher mutation rate. For instance, if a bar is differential, the bar plot is red, and PCS1 has the higher mutation rate in that comparison (the same color as bars in the bar plot of mutation rate, in that subtype).

Figure 3

Significantly different motifs in common associated genes. In total, 426 genes are associated with more than two subtypes. However, they have mutated in different motifs. (a) PTPRD and (b) ROBO2 are two oncogenes that are good examples of this phenomenon. Although they are associated with four subtypes, as it is evident in their bar plot of motif rates, there are multiple differently mutated motifs when rates in subtypes are compared. Each arrow represents a significant difference in the rate of occurrence of the motif that is pointed to, and the color of the arrow indicates the comparison that motif was significant in. The p-values can be found in Table S4.

Figure 4

Signature analysis. (a) Exposure of samples to signatures. Exposure of each sample to each signature indicates the engagement level of a sample. For example, samples of PCS5 are more exposed to signature 2 of this subtype. This indicates that the molecular mechanism associated with this signature has potentially more affected samples of this subtype. (b) Comparing deciphered signatures to COSMIC signatures. This comparison can lead to revealing associated molecular mechanisms causing PC subtype signatures. Each cell of this heatmap indicates a level of similarity.

Figure 5

Rate of mutated transcripts in subtypes. Some subtypes tend to mutate more in some transcripts of a gene. This can lead to different outcomes in subtypes. (a) gene DPP6. (b) gene CTNNA2. (c) gene TTN..

Figure 6

Expression boxplots, mutation, and motif. Expression levels of (a) KRAS and (b) DMD in five subtypes are illustrated in the form of boxplots. Some information about mutations in the genome is also provided in tables under each boxplot to represent the potential association of mutation (and their types) on expression levels.

Figure 7

Survival curves. Survival curves of the five subtypes reveal that PCS1 has the longest survival time, and PCS4 has the shortest. For detailed values see Table S8, and for pairwise comparison, survival curves see Table S9.