Prediction of DNA-binding Regions by Detecting Intramolecular Interactions with Pattern Mining

Predicting DNA-binding proteins and their binding sites attracts large attention in the study of bioinformatics and systems biology [1, 4, 5]. Sequence properties that have been employed in predicting DNA-binding sites include original sequence information, amino acid composition, evolutionary information, side chain pKa values, molecular mass of an amino acid, etc. In some studies, the predicted low-resolution structural information is also recruited [4]. This work aims to demonstrate that position conservation derived from pattern mining also provides valuable information for detecting binding regions. More specifically, it is the strength of concurrent conservation in protein sequences that points to the potential binding residues.