Enhancing Protein-ATP and Protein-ADP Binding Sites Prediction Using Supervised Instance-Transfer Learning

摘要

Protein-ATP and protein-ADP interactions are ubiquitous in a wide variety of biological processes. Accurately identifying ATP-binding and ADP-binding sites or pockets is of significant importance for both protein function analysis and drug design. Although much progress has been made, challenges remain, especially in the post-genome era where large volume of proteins without being functional annotated are quickly accumulated. In this study, we report an instance-transfer-learning-based predictor, ATP&ADPsite, to target both ATP-binding and ADP-binding residues from protein sequence and structural information. ATP&ADPsite first employs evolutionary information, predicted secondary structure, and predicted solvent accessibility to represent each residue sample. In the above feature space, a supervised instance-transfer-learning method is proposed to improve the ATP-binding/ADP-binding residues prediction by combining ATP-binding and ADP-binding proteins. Random under-sampling is lastly employed to solve the imbalanced data learning problem. Experimental results demonstrate that the proposed ATP&ADPsite achieves a better prediction performance and outperforms many existing sequence-based predictors. The ATP&ADPsite web-server is available at http://csbio.njust.edu.cn/bioinf/ATP&ADPsite.