Proprint
From DrugPedia: A Wikipedia for Drug discovery
ProPrInt (Protein - Protein Interaction Predictor)
Proprint web-server predicts physical or functional interactions between protein molecules. We have exploited the amino acid sequence-based descriptors such as amino acid composition, dipeptide composition, biochemical classes composition, pseudo-amino acid composition, Position Specific Scoring Matrix (PSSM) etc for the vector representation of protein sequences. For the web-server development we have employed only three models i)amino acid composition ii) dipeptide composition and iii) biochemical classes tripeptide composition. Support Vector Machine(SVM) is used to train the prediction models. Till now we have undertaken three different species i) Escherichia coli, ii) Saccharomyces cerevisiae and iii) Helicobacter pylori for this study.
Interaction between protein molecules is a vital biological phenomenon. Proteins do interact, physically or functionally, in order to perform life functions such as enzyme catalysis, ligand receptor binding, signal transduction, metabolic pathway, regulatory processes etc. In the present study we demonstrated that amino acid sequence based descriptors (such as dipeptide composition, biochemical class composition etc.) are better than the existing genome context features in predicting protein-protein interactions (PPIs). Moreover, our method also achieved better performance than some other sequence-based existing methods for PPI. Working datasets include Escherichia coli, Helicobacter pylori, and Saccharomyces cerevisiae interactions. Among different descriptors dipeptide composition, used for the first time to predict PPI, showed relatively better performance than other sequence-based descriptors. Support Vector Machine (SVM) is quite efficient in discriminating interacting pairs from non-interacting pairs of proteins in comparison to Artificial Neural Network (ANN). We have launched a prediction server “ProPrInt” based on the models developed in this study. ProPrInt is freely available for academia at [1].
