Improving prediction of secondary structure, local backbone angles and solvent accessible surface area of proteins by iterative deep learning

Abstract

Direct prediction of protein structure from sequence is a challenging problem. An effective approach is to break it up into independent sub-problems. These sub-problems such as prediction of protein secondary structure can then be solved independently. In a previous study, we found that an iterative use of predicted secondary structure and backbone torsion angles can further improve secondary structure and torsion angle prediction. In this study, we expand the iterative features to include solvent accessible surface area and backbone angles and dihedrals based on Cα atoms. By using a deep learning neural network in three iterations, we achieved 82% accuracy for secondary structure prediction, 0.76 for the correlation coefficient between predicted and actual solvent accessible surface area, 19° and 30° for mean absolute errors of backbone φ and ψ angles, respectively, and 8° and 32° for mean absolute errors of Cα-based θ and τ angles, respectively, for an independent test dataset of 1199 proteins. The accuracy of the method is slightly lower for 72 CASP 11 targets but much higher than those of model structures from current state-of-the-art techniques. This suggests the potentially beneficial use of these predicted properties for model assessment and ranking.

SMCHD1 mutations associated with a rare muscular dystrophy can also cause isolated arhinia and Bosma arhinia microphthalmia syndrome

“…Three predictive methods were used: NetsurfP 50 , I-TASSER 51 and SPIDER2 (ref. 52). Residues altered in both disorders (136–137) were excluded from this analysis.…”

Predicting the Functional Impact of KCNQ1 Variants of Unknown Significance

“…49, 50 Specifically, the advantages of simultaneously training a neural network to predict multiple outcome variables (disease severity, electrophysiological parameters, etc.) may enable a more accurate prediction of phenotype traits as well.…”

Genome-scale characterization of RNA tertiary structures and their functional impact by RNA solvent accessibility prediction

“…Covering the entire sequence for small RNAs by an optimized window size of 40 ensures the best learning for small RNAs while optimizing the performance for large RNAs. This window size of 40 is much larger than a typical window size (8-10) of a protein used for predicting protein secondary structure and ASA (Zhou and Faraggi 2010;Heffernan et al 2015). This is because 20 different amino acids of proteins with a typical window size of eight can generate 340 features [20×(2 × 8 + 1)], whereas four-letter-coded RNAs require us to expand the window size to 40 in order to achieve a similar number of features [324 = 4×(2 × 40 + 1)] so that a similar amount of information content can be learned from RNAs or from proteins.…”

“…To address this question, we focus on the solvent accessibility or the fraction of solvent accessible surface area (ASA) of nucleotides in the tertiary structure of an RNA chain. RNA solvent accessibility, reflecting the level of exposure of a base to solvent, is a one-dimensional structural property of tertiary structure that is more amenable for computational prediction than three-dimensional structure, as demonstrated in predicting the solvent accessibility of proteins (Zhou and Faraggi 2010;Heffernan et al 2015). The ability to predict RNA accessibility of functional tertiary structures directly from RNA sequences would support the notion that functional RNA structures are encoded in their nucleotide sequences alone (i.e., without inputting information from their binding partners).…”