CONICET | Buscador de Institutos y Recursos Humanos

Cancer treatment has been based on the pathological classification of tissue samples. However, it has been reported that tumours with the same origin can differ in genomic aberrations, aggressivity and sensitivity to drugs. To overcome this problem, classifications based on gene-level mutational patterns have been proposed. But this classification has unnoticed the effects distinct mutations can have on a given gene i.e.EGFR G735S, G796S and E804G showedoncogenicactivationin prostate cancer while R841K has no effect. Moreover, particular mutations have been found to be effective targets for cancer therapy,as mutation-specifictreatmentsavoidsystemicsideeffects.Sorafenibactsonrenal and hepaticcellcarcinomas specificallyagainstcellsbearingV600E in BRAF kinase.Somepatterns of co-occurrenceand exclusion within mutations have been reported. In this work,we propose a cancer classification based on position-level mutational patterns, by analysing COSMICv75 data and known driver mutations from kinases and Ras proteins. This allow us to suggest therapy reassignation and new therapy targets. Additionally, by estimating conditional mutational probabilities, we ve defined a network of significantly related loci. The network analysis shows that: i) pairs of driver mutations tend to exclude one another and appear recurrently in several cancer types; ii) exclusions involve same protein loci while co-mutations engage different proteins. Altogether, this analysis let us nominate new possible mutational markers and driver positions.