Protein-DNA interactions play important roles in many biological processes. On the interfaces of these protein-DNA complexes, a group of residues are identified to contribute most to structural binding free energy,they are called hot spots.

inpPDH is an improved prediction model of PrPDH which can efficiently predict hot spots as well as non-hot spots in protein-DNA interaction interfaces. To improve the prediction performance, we develop two novel feature encoding strategies and obtain 8 interfacial neighbor properties. Combined with 16 traditional features, we totally obtain 24 features for further two-step feature selection. Finally, a subset of 7 optimal features are chose to construct the predictor using support vector machine.