Classificação estrutural de famílias de proteínas com base em mas de contatos.

Abstract

The objective of this work was to verify if it is possible to classify protein chainstructures using only the chemical interactions between its residues. Through contactmaps and using three different metrics based on image processing techniques wehave showed that we are able to classify such structures in families of similar structureand function with precision up to 99%. We have performed some experiments with attributesvariation to find possible components of the structural signatures of each of thestudied protein families. We have verified that some types of interactions are more discriminatorthen others (they are hydrogen bonds without water molecules in the middleof residues, hydrophobic contacts and ion-ion linking) and that other are less discriminator(hydrogen bonds intermediated by water molecules). We also have showed thatcontacts between residues which are sequentially close (less than 30 residues of distance)are not very discriminator attributes for classification, apparently being noisesin the process. Moreover, for the preliminary results, the residues that form a greatnumber of contacts are not more important that the less connected ones as one shouldpreviously think. Residues with few contacts apparently are essential in the definitionof the structural signature of a family. We have showed that one of the techniques forcontact maps comparison can additionally be useful as an heuristic for the contact mapoverlap problem. It can be used to align contact maps and through these alignmentswe can, for example, study mutations in residues that does not affect the pattern ofcontacts. We can compare mutant and wild proteins and also, comparatively study aprotein of diverse animal species. Another important tested use of the technique is inthe discovery of a pattern of interactions between different protein chains in complexes.In assays with serine-proteases and its inhibitors, the BPTIs, we have showed that it ispossible to define a set of potentially important contacts in the binding and stabilizationof the complexes. Some of the results of this work had been implemented and areavailable, beyond this site, in the STING (http://www.cbi.cnptia.embrapa.br/SMS).We participate of the conception and implementation of three different modules: PCD(Protein Contacts Difference), TopSiMap (Topology Similarity Map) and Topologs (adata base of similar structures being overcome as base only contacts).