FC- SVM: DNA binding Proteins prediction with Average Blocks (AB) descriptors using SVM with FC feature Selection

Abstract

DNA binding proteins have important functions and roles in various biological processes, namely regulation of transcription, DNA replication, DNA packaging, DNA repairs, and DNA rearrangement. More than 135,000 atomic-level biomolecular structures from experimental results have been stored in the Protein Data Bank (PDB) database. Therefore, we need computational methods that can predict quickly and accurately the existence of DNA binding proteins. This research proposes a new method FC-SVM that combine Support Vector Machine (SVM) with F-score (FC) feature selection method to identify DNA binding protein using average block (AB) descriptor that was extracted from position specific scoring matrix (PSSM). Evaluation of the proposed method with 10 cross validations in three datasets of PDB186, PDB594 and PDB1075 shows the results of the performance 0.66, 0.72 and 0.75 resoectively.