Improved protein model quality assessment by integrating sequential and pairwise features using deep learning

Abstract

Motivation Accurately estimating protein model quality in the absence of experimental structure is not only important for model evaluation and selection but also useful for model refinement. Progress has been steadily made by introducing new features and algorithms (especially deep neural networks), but the accuracy of quality assessment (QA) is still not very satisfactory, especially local QA on hard protein targets. Results We propose a new single-model-based QA method ResNetQA for both local and global quality assessment. Our method predicts model quality by integrating sequential and pairwise features using a deep neural network composed of both 1D and 2D convolutional residual neural networks (ResNet). The 2D ResNet module extracts useful information from pairwise features such as model-derived distance maps, co-evolution information, and predicted distance potential from sequences. The 1D ResNet is used to predict local (global) model quality from sequential features and pooled pairwise information generated by 2D ResNet. Tested on the CASP12 and CASP13 datasets, our experimental results show that our method greatly outperforms existing state-of-the-art methods. Our ablation studies indicate that the 2D ResNet module and pairwise features play an important role in improving model quality assessment. Availability and implementation https://github.com/AndersJing/ResNetQA. Contact jinboxu@gmail.com Supplementary information Supplementary data are available at Bioinformatics online.

Deep Learning-Based Advances in Protein Structure Prediction

“…Xu’s group developed ResNetQA [ 69 ] (a ResNet-based QA method for both local and global quality assessment of a protein model. ResNetQA is a single-model method.…”

“…QDeep [ 68 ] uses ResNet for estimating the quality of a protein structural model. Similarly, ResNetQA [ 69 ] is a local and global quality assessment method composed of both 1D and 2D convolutional residual neural networks (ResNet). DeepAccNet [ 72 ] estimates per-residue accuracy and residue–residue distance signed error in protein models and uses these predictions to guide Rosetta protein structure refinement using a convolutional neural network.…”

Spherical convolutions on molecular graphs for protein model quality assessment

“…There have been multiple approaches developed over last 30 years. These include physics-based techniques [Randall andBaldi, 2008, Faraggi andKloczkowski, 2014], statistical and unsupervised methods, such as DFIRE [Zhou and Zhou, 2002], DOPE [Shen and Sali, 2006], GOAP [Zhou and Skolnick, 2011], RWplus [Zhang and Zhang, 2010], ORDER_AVE [Liu et al, 2014], VoroMQA [Olechnovič and Venclovas, 2014] and more, classical ML-approaches ModelEvaluator [Wang et al, 2009], ProQ2 [Ray et al, 2012], Wang_SVM [Liu et al, 2016], Qprob [Cao and Cheng, 2016], SBROD [Karasikov et al, 2019], a learning-to-rank technique [Jing et al, 2016], deep learning methods [Derevyanko et al, 2018, Conover et al, 2019, Sato and Ishida, 2019, Jing and Xu, 2020, Hiranuma et al, 2020, neural [Wallner and Elofsson, 2003], and graph neural networks [Baldassarre et al, 2020, Sanyal et al, 2020, Igashov et al, 2020.…”

ROPIUS0: A deep learning-based protocol for protein structure prediction and model selection and its performance in CASP14

“…The main difference from existing methods that use predicted distances between residues to predict model quality scores (Jing and Xu, 2020) is that ROPIUS0 uses distances predicted by the deep learning module directly for model accuracy estimation and selection. We show that the direct use of predicted distances may be sufficient to achieve satisfactory results even in the presence of some limitations, which we discuss in Section 3.…”