Deep learning methods for protein torsion angle prediction

Abstract

BackgroundDeep learning is one of the most powerful machine learning methods that has achieved the state-of-the-art performance in many domains. Since deep learning was introduced to the field of bioinformatics in 2012, it has achieved success in a number of areas such as protein residue-residue contact prediction, secondary structure prediction, and fold recognition. In this work, we developed deep learning methods to improve the prediction of torsion (dihedral) angles of proteins.ResultsWe design four different deep learning architectures to predict protein torsion angles. The architectures including deep neural network (DNN) and deep restricted Boltzmann machine (DRBN), deep recurrent neural network (DRNN) and deep recurrent restricted Boltzmann machine (DReRBM) since the protein torsion angle prediction is a sequence related problem. In addition to existing protein features, two new features (predicted residue contact number and the error distribution of torsion angles extracted from sequence fragments) are used as input to each of the four deep learning architectures to predict phi and psi angles of protein backbone. The mean absolute error (MAE) of phi and psi angles predicted by DRNN, DReRBM, DRBM and DNN is about 20–21° and 29–30° on an independent dataset. The MAE of phi angle is comparable to the existing methods, but the MAE of psi angle is 29°, 2° lower than the existing methods. On the latest CASP12 targets, our methods also achieved the performance better than or comparable to a state-of-the art method.ConclusionsOur experiment demonstrates that deep learning is a valuable method for predicting protein torsion angles. The deep recurrent network architecture performs slightly better than deep feed-forward architecture, and the predicted residue contact number and the error distribution of torsion angles extracted from sequence fragments are useful features for improving prediction accuracy.Electronic supplementary materialThe online version of this article (10.1186/s12859-017-1834-2) contains supplementary material, which is available to authorized users.

Deep Learning in the Biomedical Applications: Recent and Future Status

“…The application of deep learning in the field of the protein can be divided into two areas: protein structure prediction (PSP) ( Figure 9) and protein interaction prediction (PIP) ( Figure 10). The commonly used features in these various protein prediction problems are [133]: physicochemical properties, protein position specific scoring matrix (PSSM), solvent accessibility, secondary structure, protein disorder, contact number and the estimated probability density function of errors (difference) between true torsion angles and predicted torsion angles based on related sequence fragments.…”

“…Backbone angles prediction: Prediction of the protein backbone torsion angles (Psi and Phi) can provide important information for protein structure prediction and sequence alignment[133,[135][136][137].• Protein secondary structure prediction: Prediction of the secondary structure of the protein is an important step for the prediction of the three-dimensional structure and concentrates a large part of the scientific publications[134,[138][139][140][141][142][143][144][145][146].•Protein tertiary structure (3D) prediction: Protein tertiary structure deals with the three-dimensional protein structure and gives how the regional structures are put together in space[147][148][149][150][151][152].•Protein quality assessment (QA): In the process of protein three-dimensional structure prediction, assessing the quality of the generated models accurately is crucial. The protein structure quality assessment (QA) is an essential component in protein structure prediction and analysis[153][154][155].•Protein loop modeling and disorder prediction: Biology and medicine have a long-standing interest in computational structure prediction and modeling of proteins.…”

End-to-end differentiable learning of protein structure

“…End-to-end differentiable models replace all components of such pipelines with differentiable primitives to enable joint optimization from input to output. In contrast, use of deep learning for structure prediction has so far been restricted to individual components within a larger pipeline (Aydin et al, 2012;Gao et al, 2017;Li et al, 2017;Lyons et al, 2014), for example prediction of contact maps (Liu et al, 2017;Wang et al, 2016). This stems from the technical challenge of developing an end-to-end differentiable model that rebuilds the entire structure prediction pipeline using differentiable primitives.…”

Grid-based prediction of torsion angle probabilities of protein backbone and its application to discrimination of protein intrinsic disorder regions and selection of model structures

“…Instead of multi-state secondary structure, backbone structure of proteins can be more accurately described by continuous dihedral or rotational angles about the N-Cα bond (φ), the Cα-C bond (ψ) for single residues. A number of methods have been developed for prediction of angles in discrete states [ 8 – 11 ] or continuous values [ 6 , 12 – 17 ]. For example, ANGLOR [ 15 ] employs neural networks and support vector machine to predict φ and ψ separately.…”

“…TANGLE [ 16 ] utilizes a two-level support vector regression to predict backbone torsion angles (φ, ψ) from amino acid sequences. Li et al [ 17 ] predicted protein torsion angles using four deep learning architectures, including deep neural network (DNN), deep restricted Boltzmann machine (DRBN), deep recurrent neural network (DRNN) and deep recurrent restricted Boltzmann machine (DReRBM). Most recently, Heffernan et al [ 18 ] employed long short-term memory bidirectional recurrent neural networks that allows capture of nonlocal interactions and yielded the highest reported accuracy in angle prediction.…”