Parallel prediction of protein-protein interactions using proximal SVM
RSFDGrC'05 Proceedings of the 10th international conference on Rough Sets, Fuzzy Sets, Data Mining, and Granular Computing - Volume Part II
| Hi-index | 0.00 |
The interactions between proteins are fundamental to a broad area of biological functions. In this paper, we try to predict protein-protein interactions directly from its amino acid sequence and only one associated physicochemical feature using a Support Vector Machine (SVM). We train a SVM learning system to recognize and predict interactions using a database of known protein interactions. Each amino acid has diverse features such as hydrophobicity, polarity, charge, surface tension, etc. We select only one among these features and combine it to amino acid sequence of interacting proteins. According to the experiments, we get approximately 94% accuracy, 99% precision, and 90% recall in average when using hydrophobicity feature, which is better than the result of previous work using several features simultaneously. Therefore, we can reduce a data size and processing time to 1/n and get a better result than the previous work using n features. When using other features except hydrophobicity, experiment results show approximately 50% accuracy, which is not so good to predict interactions. This work was supported by grant NO. R01-2003-000-10860-0 from the Basic Research Program of the Korea Science & Engineering Foundation.