Sorting protein decoys by machine-learning-to-rank

Abstract

Much progress has been made in Protein structure prediction during the last few decades. As the predicted models can span a broad range of accuracy spectrum, the accuracy of quality estimation becomes one of the key elements of successful protein structure prediction. Over the past years, a number of methods have been developed to address this issue, and these methods could be roughly divided into three categories: the single-model methods, clustering-based methods and quasi single-model methods. In this study, we develop a single-model method MQAPRank based on the learning-to-rank algorithm firstly, and then implement a quasi single-model method Quasi-MQAPRank. The proposed methods are benchmarked on the 3DRobot and CASP11 dataset. The five-fold cross-validation on the 3DRobot dataset shows the proposed single model method outperforms other methods whose outputs are taken as features of the proposed method, and the quasi single-model method can further enhance the performance. On the CASP11 dataset, the proposed methods also perform well compared with other leading methods in corresponding categories. In particular, the Quasi-MQAPRank method achieves a considerable performance on the CASP11 Best150 dataset.

An overview of comparative modelling and resources dedicated to large-scale modelling of genome sequences

“…secondary-structure similarity and solvent accessibility) and statistical potential energy terms . MQAPRank is a machinelearning-to-rank method that extracts features from statistical potentials and the scores obtained from a few model-quality assessment methods (Jing et al, 2016). SVMQA is an SVM method that combines eight statistical potential energy terms and 11 consistency-based terms (between the predicted values from the sequence of the query protein and the calculated values from the model built; Manavalan & Lee, 2017).…”

“…SVMQA is an SVM method that is based on the combination of two independent predictors trained on the TM score or GDT_TS score (Manavalan & Lee, 2017). Other methods exploit deep learning and machinelearning-to-rank, which seem to be superior to SVMs (Uziela et al, 2016(Uziela et al, , 2017Jing et al, 2016).…”

Deep convolutional networks for quality assessment of protein folds

“…The problem of decoy quality assessment is essentially a ranking problem: we have to arrange decoys according to their similarity to the corresponding native structure as quantified, for instance, by the GDT TS score [38]. Such a ranking approach has recently been used by the MQAPRank method [39], which, however, relies on a support vector machine model and uses high-level features as input.…”

An Energy Landscape Treatment of Decoy Selection in Template-Free Protein Structure Prediction

“…These methods first select some high-quality structures as references and then compare the rest of the decoys with the reference structures [29]. Quasi single-model methods are shown to improve decoy selection over single-model and consensus-seeking methods [58,59].…”

“…Single-model methods assess the quality of one structure at a time [27][28][29][30] via a scoring that can be physics-based, or knowledge-based. Physics-based scoring functions model specific atomic interactions (e.g., electrostatic, hydrogen bonding, Van Der Waals interactions, and others) [31][32][33].…”