Amino acid torsion angles enable prediction of protein fold classification

Abstract Background Protein structure can provide insights that help biologists to predict and understand protein functions and interactions. However, the number of known protein structures has not kept pace with the number of protein sequences determined by high-throughput sequencing. Current techniques used to determine the structure of proteins, such as X-ray crystallography and Nuclear Magnetic Resonance (NMR) spectroscopy, are complex and may require a lot of time to analyze the experimental results, especially for large protein molecules. The limitations of these methods have motivated us to create a new approach for protein structure prediction.Results Here we describe a new approach that uses integration and analysis of torsion angle information from the Protein Data Bank to enable prediction of protein structures from amino acid sequences. Our prediction model performed well in comparison with previous methods when applied to the structural classification of two CATH datasets with more than 5000 protein domains. This new prediction model performs well with an average of 92.5% accuracy for structure classification, which is higher than the previous research. We also used our model to predict four known protein structures with a single amino acid sequence, while many other existing methods could only obtain one possible structure for a given sequence.Conclusions The results show that our method provides a new effective and reliable tool for protein structure prediction research.