engleski

Prediction of protein-protein hetero interaction sites from local sequence information using Random Forest

Identifying the interface between two interacting proteins provides important clues to the function of a protein, and is becoming increasing relevant to drug discovery. Many studies indicate that the compositions of contacting residues are unique. Here, we describe a method using random forest that identifies protein-protein interfaces from sequence. For prediction we use a non-redundant set of 333 protein complexex. The AUC for random forest classifier is 0.76. When 75% of our predictions were right, we correctly predicted 34% of all interaction sites. In almost all proteins we correctly predicted at least one interaction site. Furthermore, when in prediction we included residues that are up to 5 residues far from our predicted size, we covered 65% of all interaction sites. These results strongly indicate that prediction of interaction sites from sequence alone is possible and comparable with results obtained using structure information. Incorporating predicted structural information like ASA, secondary structure, depth and protrusion index may improve our method.

protein ; prediction ; interaction ; sequence ; random forest

nije evidentirano