Improving prediction of helix–helix packing in membrane proteins using predicted contact numbers as restraints

Abstract

One of the challenging problems in computational prediction of tertiary structure of helical membrane proteins (HMPs) is the determination of rotation of α-helices around the helix normal. Incorrect prediction of rotation of α-helices around the helix normal substantially disrupts native residue–residue contacts while inducing only a relatively small effect on the overall fold. To address this problem, we previously developed a predictor for residue contact numbers (CNs), which measure the local packing density of residues within the protein tertiary structure. In this study, we tested the idea of incorporating predicted CNs as restraints to guide the sampling of helix rotation. For a benchmark set of 15 HMPs with simple to rather complicated folds, the average contact recovery (CR) of best-sampled models was improved for all targets, the likelihood of sampling models with CR greater than 20% was increased for 13 targets, and the average RMSD100 of best-sampled models was improved for 12 targets. This study demonstrated that explicit incorporation of CNs as restraints improves the prediction of helix-helix packing.

Interfaces Between Alpha-helical Integral Membrane Proteins: Characterization, Prediction, and Docking

“…Previously we demonstrated that residue WCNs [23] and other experimental or simulated restraints [94] can be effectively incorporated into scoring functions to improve de novo tertiary structure prediction for α-helical membrane proteins. Likewise, we hypothesized that a scoring term that assigns a penalty score to docked models according the magnitude of the deviation the WCNs of predicted interface residues from their predicted WCNs will also be highly effective:italicPenalty scoregoodbreak=1n∑i∈italicpredicted0.25emitalicinterface0.25emitalicresiduesnitalicWCNi−italicWCNip2where WCN i is the WCN of interface residue i computed based on the docked model, WCN i p is the WCN of interface residue i predicted by the neural network.…”

“…On the one hand, computing residue WCN is computationally less demanding than computing residue solvent accessibility. This makes it feasible to compute residue WCNs on the fly and use them directly for scoring in protein structure prediction and protein-protein docking predictions where an enormous amount of conformational space must be sampled [23]. On the other hand, we had previously developed a machine-learning framework that allowed accurate prediction of residue WCNs for α-helical IMPs [68].…”

“…WCN is defined as the number of neighboring residues weighted by their proximity to the focal residue, it measures the local packing degree of residues within the protein tertiary structure [21,22]. We find that interface residues can be accurately identified based on the deviation of predicted WCN from the WCN computed from the protomer structure alone.Structure-based docking of IMPs: Based on our previous findings in which we demonstrated that residue WCNs can be effectively used as restraints to improve de novo tertiary structure prediction for α-helical IMPs [23], we implemented an algorithm which leverages the high discriminatory power of a WCN-based penalty score for accurate docking of α-helical IMPs.…”

“…Structure-based docking of IMPs: Based on our previous findings in which we demonstrated that residue WCNs can be effectively used as restraints to improve de novo tertiary structure prediction for α-helical IMPs [23], we implemented an algorithm which leverages the high discriminatory power of a WCN-based penalty score for accurate docking of α-helical IMPs.…”

Predicting the Functional Impact of KCNQ1 Variants of Unknown Significance

“…We chose a neural network, which generally is considered to be a nonlinear model, for the present study to leverage our extensive experience with neural networks and an established library for feature engineering and model building. 42–46 Admittedly, a logistic regression model performed only slightly worse (AUC = 0.855) and a linear discriminant classifier performed comparably (AUC = 0.870). However, given the complexity in the mechanisms behind KCNQ1 dysfunction, we expect that the “true” decision boundary between normal and dysfunctional variants is complex.…”