Usando RINs para entender as mutações em câncer: mutações deletérias são mais comumente associadas a aminoácidos altamente conectados

Abstract

In the last decades, advances in whole­genome sequencing research lead to the identification of a vast number of cancer­related mutations. Achieving high performance in estimating the impacts of cancer mutations on protein structure is not an easy task, and most studies are limited to one­by­one whole structural analysis. Moreover, there are still many challenges on the way to the precise and automated prediction of deleterious mutations. Therefore, understanding the structural impact of a particular amino acid change is hugely important for cancer medical research. However, most studies have been emphasizing sequences and structural modifications based on chemical characteristics of amino acids, not in fold features in which the conservation of non­covalent interactions play a significant role. Henceforth, in the present study, we used residue interaction networks (RINs) for large­scale analysis of cancer missense mutations in order to infer their effects on the conservation of non­covalent interactions. We hypothesize that changes in highly connected amino acids are more likely to cause deleterious mutations. To evaluate this, we retrieved cancer missense mutations from COSMIC (cancer.sanger.ac.uk/cosmic) and TCGA (cancergenome.nih.gov) databases and mapped them to their respective structures retrieved from Protein Data Bank (rcsb.org). Then, RINs were constructed from the obtained pdb files, and network parameters such as the node's degree, edges' type, clustering coefficient, betweenness weighted were assessed and plotted using R scripts. Later, we compared these results against reported missense single nucleotide polymorphisms retrieved from dbSNP (www.ncbi.nlm.nih.gov/projects/SNP/) and to pathogenic and non­pathogenic cancer mutations from ClinVar (www.ncbi.nlm.nih.gov/clinvar/) databases. Our results demonstrate that the distribution of mutations per degree (node connectivity) varies significantly compared to random Monte Carlo simulations, tending to remain at nodes with lower connectivity. We also compare with the distribution of a set of human single nucleotide polymorphisms (SNPs). Besides, the proportion of deleterious mutations was significantly increased in nodes with a high degree of connectivity when two different criteria were used for their classification: proportions of software predictors (Ndamage) and clinical classification obtained from ClinVar. Considering these results, we can conclude that the changes in the highly connected amino acids are, in fact, more prone to generate deleterious mutations, due their higher proportion of occurrence in these nodes. Our results also indicate that the conservation of non­covalent interactions is an important parameter to consider in the evaluation of mutations effects and RINs analysis can be used as an additional parameter to aid in the prediction of deleterious mutations in cancer.