Software for Human Protein-Protein Interaction Prediction from Multiple Sources Integration
This directory contains the supplementary files for paper:
Yanjun Qi[1]1, Harpreet K. Dhiman2, Neil Bhola3, Ivan Budyak4, Siddhartha Kar5, David Man2, Arpana Dutta2, Kalyan Tirupula2, Brian I. Carr5, Jennifer Grandis3, Ziv Bar-Joseph1ß and Judith Klein-Seetharaman1,2,4
Title: Systematic prediction of human membrane receptor interactions, PROTEOMICS (2009)
Supplementary Information see: http://www.cs.cmu.edu/~qyj/HMRI/
You can also download a tar version of this software @ http://www.cs.cmu.edu/afs/cs.cmu.edu/project/structure-9/PPI/HMRI/software/
-----------------------------------------------
HPPIPM is the name of a software aiming for predicting human protein-protein intearctions by integrating multiple biological data sources based on "Random Forest" classifier. It provides a java based GUI interface and perl based interface from command line.
Human membrane receptor interactome is provided as a test case for this software.
@article{qi2009systematic, title={Systematic prediction of human membrane receptor interactions}, author={Qi, Yanjun and Dhiman, Harpreet K and Bhola, Neil and Budyak, Ivan and Kar, Siddhartha and Man, David and Dutta, Arpana and Tirupula, Kalyan and Carr, Brian I and Grandis, Jennifer and others}, journal={Proteomics}, volume={9}, number={23}, pages={5243--5255}, year={2009}, publisher={Wiley Online Library} }
Abstract
Membrane receptor-activated signal transduction pathways are integral to cellular functions and disease mechanisms in humans. Identification of the full set of proteins interacting with membrane receptors by high-throughput experimental means is difficult because methods to directly identify protein interactions are largely not applicable to membrane proteins. Unlike prior approaches that attempted to predict the global human interactome, we used a computational strategy that only focused on discovering the interacting partners of human membrane receptors leading to improved results for these proteins. We predict specific interactions based on statistical integration of biological data containing highly informative direct and indirect evidences together with feedback from experts. The predicted membrane receptor interactome provides a system-wide view, and generates new biological hypotheses regarding interactions between membrane receptors and other proteins. We have experimentally validated a number of these interactions. The results suggest that a framework of systematically integrating computational predictions, global analyses, biological experimentation and expert feedback is a feasible strategy to study the human membrane receptor interactome.