Accurate De Novo Prediction of Protein Contact Map by Ultra-Deep Learning Model

Abstract

MotivationProtein contacts contain key information for the understanding of protein structure and function and thus, contact prediction from sequence is an important problem. Recently exciting progress has been made on this problem, but the predicted contacts for proteins without many sequence homologs is still of low quality and not extremely useful for de novo structure prediction.\n\nMethodThis paper presents a new deep learning method that predicts contacts by integrating both evolutionary coupling (EC) and sequence conservation information through an ultra-deep neural network formed by two deep residual neural networks. The first residual network conducts a series of 1-dimensional convolutional transformation of sequential features; the second residual network conducts a series of 2-dimensional convolutional transformation of pairwise information including output of the first residual network, EC information and pairwise potential. By using very deep residual networks, we can model contact occurring patterns and very complex sequence-structure relationship and thus, obtain high-quality contact prediction regardless of how many sequence homologs are available for proteins in question.\n\nResultsOur method greatly outperforms existing methods and leads to much more accurate contact-assisted folding. Tested on 105 CASP11 targets, 76 past CAMEO hard targets, and 398 membrane proteins, the average top L long-range prediction accuracy obtained our method, one representative EC method CCMpred and the CASP11 winner MetaPSICOV is 0.47, 0.21 and 0.30, respectively; the average top L/10 long-range accuracy of our method, CCMpred and MetaPSICOV is 0.77, 0.47 and 0.59, respectively. Ab initio folding using our predicted contacts as restraints but without any force fields can yield correct folds (i.e., TMscore>0.6) for 203 of the 579 test proteins, while that using MetaPSICOV- and CCMpred-predicted contacts can do so for only 79 and 62 of them, respectively. Our contact-assisted models also have much better quality than template-based models especially for membrane proteins. The 3D models built from our contact prediction have TMscore>0.5 for 208 of the 398 membrane proteins, while those from homology modeling have TMscore>0.5 for only 10 of them. Further, even if trained by only non-membrane proteins, our deep learning method works very well on membrane protein contact prediction. In the recent blind CAMEO benchmark, our fully-automated web server implementing this method successfully folded 4 targets with a new fold and only 0.3L-2.3L effective sequence homologs, including one {beta} protein of 182 residues, one +{beta} protein of 125 residues, one protein of 140 residues and one protein of 217 residues.\n\nAvailability: http://raptorx.uchicago.edu/ContactMap/\n\nAuthor SummaryProtein contact prediction from sequence alone is an important problem. Recently exciting progress has been made on this problem due to the development of direct evolutionary coupling analysis (DCA). However, DCA is effective on only some proteins with a very large number (>1000) of sequence homologs. To further improve contact prediction, we borrow ideas from the latest breakthrough of deep learning, a powerful machine learning technique that has recently revolutionized object recognition, speech recognition and the GO game. We have developed a new deep learning method that predicts contacts by integrating both sequence conservation and co-variation information through an ultra-deep neural network, which can model very complex relationship between sequence and contact map as well as high-order correlation among residues.\n\nOur test results suggest that deep learning can revolutionize protein contact prediction. Tested on 398 membrane proteins, the L/10 long-range accuracy obtained by our method is 77.6% while that by the state-of-the-art methods CCMpred and MetaPSICOV is 51.8% and 61.2%, respectively. Ab initio folding using our predicted contacts as restraints can generate much better 3D structural models than the other contact prediction methods. In particular, without using any force fields our predicted contacts yield correct folds for 203 of the 579 test proteins, while MetaPSICOV- and CCMpred can do so for only 79 and 62 of them, respectively. Our contact-assisted models also have much better quality than template-based models (TBM) built from the training proteins. For example, our contact-assisted models have TMscore>0.5 for 208 of the 398 membrane proteins while TBMs have TMscore >0.5 for only 10 of them. Even without using any membrane proteins to train our deep learning models, our method still performs very well on membrane protein contact prediction. Recent blind test of our method in CAMEO shows that our method successfully folded 4 targets with a new fold and only 0.3L-2.3L effective sequence homologs.