MDCinfer

Analysis on Multi-domain Cooperation for Predicting
Protein-Protein Interactions

 [Version 1.0]

Aug.26, 2007

http://intelligent.eic.osaka-sandai.ac.jp/chenen/MDCinfer.htm

　

AUTHORS
=======

Rui-Sheng Wang(wangrsh@amss.ac.cn) Yong Wang(ywang@amss.ac.cn)

Ling-Yun Wu(lywu@amt.ac.cn)

Xiang-Sun Zhang (zxs@amt.ac.cn) Luonan Chen (chen@elec.osaka-sandai.ac.jp)

METHOD
======

MDCinfer aims to infer protein-protein interaction by considering cooperative domain interactions. Unlike most existing methods, it assumes cooperative-domain pairs but not only single-domain pairs as the basic units of a protein interaction. The interaction probabilities of single-domain pairs and cooperative-domain pairs are computed by a linear programming algorithm and a fast association probabilistic method. Novel protein interactions can be predicted by an extended probabilities model which can accommodate cooperative-domain pairs.

PROCEDURE
=========

MDCinfer.exe: Filename1 Filename2 Filename3 Negnum Methstr

This software is to predict novel protein-proteins interactions from training protein interactions, where Filename1 is an input file containing protein interactions which will act as a training set. Its format can be referred to an example called train.txt in MDCinfer.rar and Filename2 is another input file containing the test set, i.e. the protein pairs whose interaction states you want to know based on the training set. Also its format can be referred to an example called test.txt in MDCinfer.rar. Filename3 is a file containing the protein-domain relationships for the proteins in filename2 and its format can be referred to a file called testdomain.txt in MDCinfer.rar. Negnum is an integer number and indicates how many negative samples you want to add during training.  Methstr is a string denoting which method you want to choose. It can only take two values: “APMM”denotes the fast association probabilistic method based on multi-domain cooperation, and “LPM”denotes the linear programming algorithm based on multi-domain cooperation.

 

REFERENCE
=========

Analysis on Multi-domain Cooperation for Predicting Protein-Protein Interactions. (In submission)

 

SOFTWARE
========

This is a beta version of the program for preliminary testing. The program is still under development.

MDCinfer.rar

 

SUPPLEMENTARY MATERIALS
========

Numeriacal PPI data: Ito et al., 2001, Krogan et al.,2006

Binary PPI data: (From Liu et al.,2005) yeast, worm,fly, domain information

PPI data of multiple species in DIP: (From Riely et al., 2005) PPI data, Proteins, Domains

Predicted DDIs on DIP data: APMM(1) APMM(2)

Overlap with iPfam (top 3005) APMM(1) APMM(2)

All the cooperative-domain interactions in MIPS

All the cooperative-domain interactions in DIP

All the cooperative-domain interactions in Krogan data set

All the cooperative-domain interactions verified by complex in PDB