Recognition of Protein Interaction Regions Through Time-Frequency Analysis

Protein–protein interactions address most of the molecular processes in all nature, even also infection driven by pathogens can be explained by interactions between proteins. Pathogens, like virus, bacteria, and parasites express proteins in their surfaces that help the pathogens to invade the target host cells. Thus, it is pivotal to know which regions in pathogenic proteins interact with host cell receptors. In these days, meaningful pathogen databases are available, and many pathogenic proteins have been well studied; but many other pathogenic proteins have not been characterized yet. This work applied Time-Frequency Analysis (TFA) to recognize important regions in proteins. TFA shows the highest local variances in a protein string from three different time-frequency distributions. We sought to know if this approach is able to recognize stretched residues related to interaction. Our approach was applied in some study cases from pathogenic co-crystallized structures. We searched the frequency/variance that characterizes interaction regions in pathogenic proteins and with this information tried to identify new interaction regions in either paralogs or orthologs. We found that this approach was able to detect under several descriptors important regions in proteins even those related to interaction. To analyze the performance of TFA detecting interaction regions a confusion matrix was performed from 127 proteins and we obtained an acceptable sensitivity and specificity for some descriptors (sensitivity/specificity around 0.85). We identified a peptide from a mouse protein IRGb2-b1 which showed the highest local variance in TFA model and this peptide was assessed in a growth assay in Toxoplasma gondii model. The peptide was able to delay Toxoplasma growing.