Fig. 1

An overview of our proposed pipeline. Here we first downloaded publicly available somatic point mutations of 536 whole genome sequencing CRC individuals from the ICGC. We then used the FANTOMCAT genes list to identify the number of mutations in coding and lncRNA genes. After that, we used negative-binomial and beta-binomial distributions to identify significantly mutated genes and gene-motifs, respectively. Using 3131 candidate gene-motifs as features of our model-based clustering, we identified seven CRC subtypes. Our comprehensive biological analysis showed that the identified subtypes have different mutational load in different genes. Our mutational signature analysis also showed that different combinations of signatures can be observed in each subtype. We also identified genes and conserved motifs that significantly mutated in each subtype. Finally, we performed gene ontology, pathway, and survival curve analyses