Quick Reliable Exploration of the PDB Universe Seeks a New Template


Quick Reliable Exploration of the PDB Universe Seeks a New Template

Near-native protein structure prediction through Template Based Modelling (TBM) has been a major realistic goal of structural biology for several years. The TBM algorithms require the best-set of templates for a target protein sequence to maximally cover it and construct its correct topology. However, the accuracy of such prediction algorithms suffers from the algorithmic and logical problems of our template search measures which fail to quickly screen reliable structures for a target sequence. In this study, we employ the culled PDB95 dataset of 41,967 templates to predict the CASP10 target T0752 models for assessing the efficiency of the usually employ search engines PSI-BLAST and HHPred. Our analysis presents a detailed study in order to open new vistas for improving the accuracy of TBM prediction methodologies. It reveals weaknesses of most popular template search measures and thereby briefly provides a significant insight into the qualities of a foreseen template search algorithm to illustrate the need for a more reliable template search algorithm.

Because protein modelling accuracy is primarily rooted in the structural and functional homology of template against the target sequence, a thorough template search tool is mandatorily required to screen all the reliable hits for predicting the most accurate target models consistently. In this study, the template search measures PSI-BLAST and HHPred are scrutinized to assess the entirety of their results. Among a culled PDB95dataset of 41,967 structures 96,920 PDB entries in total and out of 24 templates with TM_Score more than 0.7, PSI-BLAST and HHPred protocols fail to search 23 and 22 hits, respectively. Therefore, this study attempts to highlight the logical problems prevailing in our normally employed template search measures and bridge the path for our further research methodologies.

Journal of Data Mining in Genomics & Proteomics accept Editorial, letter, case reports, short communications, original and review articles from leading scientists and scholars around the world in all areas of food, nutrition, biochemistry, physiology, pathology and medicine.

Authors can directly submit the article as an attachment to this online: https://www.longdom.org/submissions/data-mining-genomics-proteomics.html

The articles once published can be accessed freely.

Media Contact:
Alex John
Journal Manager
Journal of Data Mining in Genomics & Proteomics
Email: proteomdataming@oajournal.org