Knowledge base and neural network approach for protein secondary structure prediction

•A novel neuro-statistical algorithm using knowledge base and neural network for PSSP is proposed.•Association of 5-residue word with corresponding secondary structure forms the knowledge base.•Lateral and hierarchical validation is employed for PSSP.•A Backpropogation neural network is used to model the exceptions in the knowledge base.•The Q3 accuracy of 90% and 82% is achieved on the RS126 and CB396 test data sets respectively.

Protein structure prediction is of great relevance given the abundant genomic and proteomic data generated by the genome sequencing projects. Protein secondary structure prediction is addressed as a sub task in determining the protein tertiary structure and function. In this paper, a novel algorithm, KB-PROSSP-NN, which is a combination of knowledge base and modeling of the exceptions in the knowledge base using neural networks for protein secondary structure prediction (PSSP), is proposed. The knowledge base is derived from a proteomic sequence–structure database and consists of the statistics of association between the 5-residue words and corresponding secondary structure. The predicted results obtained using knowledge base are refined with a Backpropogation neural network algorithm. Neural net models the exceptions of the knowledge base. The Q3 accuracy of 90% and 82% is achieved on the RS126 and CB396 test sets respectively which suggest improvement over existing state of art methods.

Graphical abstractFigure optionsDownload full-size imageDownload as PowerPoint slide