2017/01/27 by Dengming Ming, Min Han, Ming, Dengming +3
Biochemistry, Genetics and Molecular Biology · Computer Science · #92-08 #Biomolecules (q-bio.BM) #Computational Drug Discovery Methods #FOS: Biological sciences #J.3 #Machine Learning in Bioinformatics #Protein Structure and Dynamics
paper · pdf · doi:10.48550/arxiv.1701.08086
openalex publication_date 2017/01/27 · openalex created_date 2017/02/10 · openalex updated_date 2026/07/28
Identifying protein functional sites (PFSs) and protein-ligand interactions (PLIs) are critically important in understanding the protein function and the involved biochemical reactions. As large amount of unknown proteins are quickly accumulated in this post-genome era, an urgent task arises to predict PFSs and PLIs at residual level. Nowadays many knowledge-based methods have been well developed for prediction of PFSs, however, accurate methods for PLI prediction are still lacking. In this study, we have presented a new method for prediction of PLIs and PFSs based on sequence of the inquiry protein. The key of the method hinges on a function- and interaction-annotated protein domain profile database, called fiDPD, which was built from the Structural Classification of Proteins (SCOP) database, using a hidden Markov model program. The method was applied to 13 target proteins from the recent Critical Assessment of Structure Prediction (CASP10/11). Our calculations gave a Matthews correlation coefficient (MCC) value of 0.66 for prediction of PFSs, and an 80% recall in prediction of PLIs. Our method reveals that PLIs are conserved during the evolution of proteins, and they can be reliably predicted from fiDPD. fiDPD can be used as a complement to existent bioinformatics tools for protein function annotation.