Systematic identification and quantification of substrate specificity determinants in human protein kinases = Identificación y cuantificación sistemática de determinantes de la especificidad por sustrato en las proteínas quinasas de humano

In human there are 518 protein kinases reported and many of them are involved in several cellular processes and also in important pathologies. Kinases have a diverse specificity for the sequences they phosphorylate, and it has been shown that their in vivo specificity is guided by several elements such as the sequence surrounding the phosphorylated amino acid in the substrate, the co-localization with the substrate, interactions mediated by docking sites and the association of kinases to adaptor or scaffold proteins (AS). The objective of the current thesis has been the identification and quantification of the contribution to the specificity of i) the phosphorylated site and its surrounding residues in sequence and ii) the association of kinases to AS proteins. By integrating data from public resources we compiled a set of kinase-phosphorylation sites in human corresponding to 325 (62.7%) kinases, 1856 substrates and 5946 phosphorylation sites. We have used sequence logos for representing the phosphorylation motifs recognized by the kinases in our set and we have used these logos to guide the classification of kinases attending to the residue composition of the sequences surrounding the phosphorylation site. We have used position-specific scoring matrices (PSSMs) as the probabilistic representation of the sequences phosphorylated by the kinases. Based on the score in the PSSMs relative to the phospho-acceptor amino acid, we classified several residues as specificity-determinant residues (SDR) for several kinase families. The identity, position in the sequence alignment and the frequency of the SDRs identified vary considerably among the kinase families analyzed. The statistical significance of the PSSMs was assessed taking into account their recall and their information content (IC). We found negative correlations between the the number of seed sequences and i) the recall of the PSSMs and ii) the IC of the PSSMs. Based on the IC value, and in the comparison to random distributions, we found that statistically significant PSSMs differ from the non-statistically significant ones regarding their IC, recall, number of seed phosphorylation sites and AUC-ROC. We have developed a computational strategy for the identification of proteins with known function as adaptors or scaffolds of human kinases (kAS). In total, we have identified a set of 191 kAS proteins which is enriched in domains and functional terms that support their role as AS. These 191 kAS associate to 55% of human kinases, which suggest that the association to this type of AS molecules is common among human kinases. Our results suggest that, when compared to random proteins, kAS proteins are five times more likely to interact with a significantly large fraction of the substrates of a the kinases to which the kAS are associated. Starting from a set of 156 human kinases for which we count with at least five substrates, we identified a set of 279 proteins with a potential function as adaptor or scaffold (pAS). This set is enriched in protein domains and functional terms related to the predicted function and that also suggest a relationship to signalling processes. Our analysis on cellular co-localization suggest that, for 74.6% of the kinase-pAS associations found, the pAS protein may play a role in the co-localization of the kinases and their corresponding sets of substrates. Finally, we have analyzed the relationship between the association of different kinases to common AS proteins and the number of in vivo substrates shared by the kinases. Our results suggest that kinases with AS proteins in common do not share more in vivo substrates than what would be expected only due to chance. To our opinion, this suggests that AS proteins may diminish the substrate crossed specificity of the kinases to which they associate