Send to

Choose Destination
See comment in PubMed Commons below
Bioinformatics. 2003 Oct;19 Suppl 2:ii58-65.

Inferring strengths of protein-protein interactions from experimental data using linear programming.

Author information

  • 1Bioinformatics Center, Institute for Chemical Research, Kyoto University, Gokasho, Uji, Japan.



Several computational methods have been proposed for inference of protein-protein interactions. Most of the existing methods assume that protein-protein interaction data are given as binary data (i.e. whether or not each protein pair interacts). However, multiple biological experiments are performed for the same protein pairs and thus the ratio (strength) of the number of observed interactions to the number of experiments is available for each protein pair.


We propose a new method for inference of protein-protein interactions from such experimental data. This method tries to minimize the errors between the ratios of observed interactions and the predicted probabilities in training data, where this problem is formalized as a linear program based on a probabilistic model. We compared the proposed method with the association method, the EM method and the SVM-based method using real interaction data. It is shown that a variant of the method is comparable to existing methods for binary data. It is also shown that the method outperforms existing methods for numerical data.


Programs transforming input data into LP format files are available upon request.

[PubMed - indexed for MEDLINE]
Free full text
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for HighWire
    Loading ...
    Support Center