Real-value and confidence prediction of protein backbone dihedral angles through a hybrid method of clustering and deep learning

摘要

Background. Protein dihedral angles provide a detailed description of proteinlocal conformation. Predicted dihedral angles can be used to narrow down theconformational space of the whole polypeptide chain significantly, thus aidingprotein tertiary structure prediction. However, direct angle prediction fromsequence alone is challenging. Method. In this study, we present a novel method to predict real-valuedangles by combining clustering and deep learning. That is, we first generatecertain clusters of angles (each assigned a label) and then apply a deepresidual neural network to predict the label posterior probability. Finally, weoutput real-valued prediction by a mixture of the clusters with their predictedprobabilities. At the same time, we also estimate the bound of the predictionerrors at each residue from the predicted label probabilities. Result. In this article, we present a novel method (named RaptorX-Angle) topredict real-valued angles by combining clustering and deep learning. Tested ona subset of PDB25 and the targets in the latest two Critical Assessment ofprotein Structure Prediction (CASP), our method outperforms the existingstate-of-art method SPIDER2 in terms of Pearson Correlation Coefficient (PCC)and Mean Absolute Error (MAE). Our result also shows approximately linearrelationship between the real prediction errors and our estimated bounds. Thatis, the real prediction error can be well approximated by our estimated bounds. Conclusions. Our study provides an alternative and more accurate predictionof dihedral angles, which may facilitate protein structure prediction andfunctional study.