Improving protein tertiary structure prediction by deep learning and distance prediction in CASP14

Substantial progresses in protein structure prediction have been made by utilizing deep-learning and residue-residue distance prediction since CASP13. Inspired by the advances, we improve our CASP14 MULTICOM protein structure prediction system by incorporating three new components: (a) a new deep learning-based protein inter-residue distance predictor to improve template-free (ab initio) tertiary structure prediction, (b) an enhanced template-based tertiary structure prediction method, and (c) distance-based model quality assessment methods empowered by deep learning. In the 2020 CASP14 experiment, MULTICOM predictor was ranked seventh out of 146 predictors in tertiary structure prediction and ranked third out of 136 predictors in inter-domain structure prediction. The results demonstrate that the template-free modeling based on deep learning and residue-residue distance prediction can predict the correct topology for almost all template-based modeling targets and a majority of hard targets (template-free targets or targets whose templates cannot be recognized), which is a significant improvement over the CASP13 MULTICOM predictor. Moreover, the template-free modeling performs better than the template-based modeling on not only hard targets but also the targets that have homologous templates. The performance of the template-free modeling largely depends on the accuracy of distance prediction closely related to the quality of multiple sequence alignments. The structural model quality assessment works well on targets for which enough good models can be predicted, but it may perform poorly when only a few good models are predicted for a hard target and the distribution of model quality scores is highly skewed. MULTICOM is available at https://github.com/jianlin-cheng/MULTICOM_Human_CASP14/tree/CASP14_DeepRank3 and https://github.com/multicom-toolbox/multicom/tree/multicom_v2.0 .     

Deep-learning contact-map guided protein structure prediction in CASP13

We report the results of two fully automated structure prediction pipelines, “Zhang-Server” and “QUARK”, in CASP13. The pipelines were built upon the C-I-TASSER and C-QUARK programs, which in turn are based on I-TASSER and QUARK but with three new modules: (a) a novel multiple sequence alignment (MSA) generation protocol to construct deep sequence-profiles for contact prediction; (b) an improved meta-method, NeBcon, which combines multiple contact predictors, including ResPRE that predicts contact-maps by coupling precision-matrices with deep residual convolutional neural-networks; and (c) an optimized contact potential to guide structure assembly simulations. For 50 CASP13 FM domains that lacked homologous templates, average TM-scores of the first models produced by C-I-TASSER and C-QUARK were 28% and 56% higher than those constructed by I-TASSER and QUARK, respectively. For the first time, contact-map predictions demonstrated usefulness on TBM domains with close homologous templates, where TM-scores of C-I-TASSER models were significantly higher than those of I-TASSER models with a P-value <.05. Detailed data analyses showed that the success of C-I-TASSER and C-QUARK was mainly due to the increased accuracy of deep-learning-based contact-maps, as well as the careful balance between sequence-based contact restraints, threading templates, and generic knowledge-based potentials. Nevertheless, challenges still remain for predicting quaternary structure of multi-domain proteins, due to the difficulties in domain partitioning and domain reassembly. In addition, contact prediction in terminal regions was often unsatisfactory due to the sparsity of MSAs. Development of new contact-based domain partitioning and assembly methods and training contact models on sparse MSAs may help address these issues.     

Assessment of protein assembly prediction in CASP13

We present the assembly category assessment in the 13th edition of the CASP community-wide experiment. For the second time, protein assemblies constitute an independent assessment category. Compared to the last edition we see a clear uptake in participation, more oligomeric targets released, and consistent, albeit modest, improvement of the predictions quality. Looking at the tertiary structure predictions, we observe that ignoring the oligomeric state of the targets hinders modeling success. We also note that some contact prediction groups successfully predicted homomeric interfacial contacts, though it appears that these predictions were not used for assembly modeling. Homology modeling with sizeable human intervention appears to form the basis of the assembly prediction techniques in this round of CASP. Future developments should see more integrated approaches where subunits are modeled in the context of the assemblies they form.     

Effect of using suboptimal alignments in template-based protein structure prediction

Computational protein structure prediction remains a challenging task in protein bioinformatics. In the recent years, the importance of template-based structure prediction is increasing because of the growing number of protein structures solved by the structural genomics projects. To capitalize the significant efforts and investments paid on the structural genomics projects, it is urgent to establish effective ways to use the solved structures as templates by developing methods for exploiting remotely related proteins that cannot be simply identified by homology. In this work, we examine the effect of using suboptimal alignments in template-based protein structure prediction. We showed that suboptimal alignments are often more accurate than the optimal one, and such accurate suboptimal alignments can occur even at a very low rank of the alignment score. Suboptimal alignments contain a significant number of correct amino acid residue contacts. Moreover, suboptimal alignments can improve template-based models when used as input to Modeller. Finally, we use suboptimal alignments for handling a contact potential in a probabilistic way in a threading program, SUPRB. The probabilistic contacts strategy outperforms the partly thawed approach, which only uses the optimal alignment in defining residue contacts, and also the re-ranking strategy, which uses the contact potential in re-ranking alignments. The comparison with existing methods in the template-recognition test shows that SUPRB is very competitive and outperforms existing methods.     

Analysis of distance-based protein structure prediction by deep learning in CASP13

This paper reports the CASP13 results of distance-based contact prediction, threading, and folding methods implemented in three RaptorX servers, which are built upon the powerful deep convolutional residual neural network (ResNet) method initiated by us for contact prediction in CASP12. On the 32 CASP13 FM (free-modeling) targets with a median multiple sequence alignment (MSA) depth of 36, RaptorX yielded the best contact prediction among 46 groups and almost the best 3D structure modeling among all server groups without time-consuming conformation sampling. In particular, RaptorX achieved top L/5, L/2, and L long-range contact precision of 70%, 58%, and 45%, respectively, and predicted correct folds (TMscore > 0.5) for 18 of 32 targets. Further, RaptorX predicted correct folds for all FM targets with >300 residues (T0950-D1, T0969-D1, and T1000-D2) and generated the best 3D models for T0950-D1 and T0969-D1 among all groups. This CASP13 test confirms our previous findings: (a) predicted distance is more useful than contacts for both template-based and free modeling; and (b) structure modeling may be improved by integrating template and coevolutionary information via deep learning. This paper will discuss progress we have made since CASP12, the strength and weakness of our methods, and why deep learning performed much better in CASP13.     

CASP13 target classification into tertiary structure prediction categories

Protein target structures for the Critical Assessment of Structure Prediction round 13 (CASP13) were split into evaluation units (EUs) based on their structural domains, the domain organization of available templates, and the performance of servers on whole targets compared to split target domains. Eighty targets were split into 112 EUs. The EUs were classified into categories suitable for assessment of high accuracy modeling (or template-based modeling [TBM]) and topology (or free modeling [FM]) based on target difficulty. Assignment into assessment categories considered the following criteria: (a) the evolutionary relationship of target domains to existing fold space as defined by the Evolutionary Classification of Protein Domains (ECOD) database; (b) the clustering of target domains using eight objective sequence, structure, and performance measures; and (c) the placement of target domains in a scatter plot of target difficulty against server performance used in the previous CASP. Generally, target domains with good server predictions had close template homologs and were classified as TBM. Alternately, targets with poor server predictions represent a mixture of fast evolving homologs, structure analogs, and new folds, and were classified as FM or FM/TBM overlap.     

A coarse-grained Langevin molecular dynamics approach to de novo protein structure prediction

De novo prediction of protein structures, the prediction of structures from amino acid sequences which are not similar to those of hitherto resolved structures, has been one of the major challenges in molecular biophysics. In this paper, we develop a new method of de novo prediction, which combines the fragment assembly method and the simulation of physical folding process: structures which have consistently assembled fragments are dynamically searched by Langevin molecular dynamics of conformational change. The benchmarking test shows that the prediction is improved when the candidate structures are cross-checked by an empirically derived score function.     

Computational methods in protein structure prediction

This review presents the advances in protein structure prediction from the computational methods perspective. The approaches are classified into four major categories: comparative modeling, fold recognition, first principles methods that employ database information, and first principles methods without database information. Important advances along with current limitations and challenges are presented.     

Evaluation of current techniques for Ab initio protein structure prediction

The results of a protein structure prediction contest are reviewed. Twelve different groups entered predictions on 14 proteins of known sequence whose structures had been determined but not yet disseminated to the scientific community. Thus, these represent true tests of the current state of structure prediction methodologies. From this work, it is clear that accurate tertiary structure prediction is not yet possible. However, protein fold and motif prediction are possible when the motif is recognizably similar to another known structure. Internal symmetry and the information inherent in an aligned family of homologous sequences facilitate predictive efforts. Novel folds remain a major challenge for prediction efforts. © 1995 Wiley-Liss, Inc.     

Protein structure prediction assisted with sparse NMR data in CASP13

CASP13 has investigated the impact of sparse NMR data on the accuracy of protein structure prediction. NOESY and ¹⁵N-¹H residual dipolar coupling data, typical of that obtained for ¹⁵N,¹³C-enriched, perdeuterated proteins up to about 40 kDa, were simulated for 11 CASP13 targets ranging in size from 80 to 326 residues. For several targets, two prediction groups generated models that are more accurate than those produced using baseline methods. Real NMR data collected for a de novo designed protein were also provided to predictors, including one data set in which only backbone resonance assignments were available. Some NMR-assisted prediction groups also did very well with these data. CASP13 also assessed whether incorporation of sparse NMR data improves the accuracy of protein structure prediction relative to nonassisted regular methods. In most cases, incorporation of sparse, noisy NMR data results in models with higher accuracy. The best NMR-assisted models were also compared with the best regular predictions of any CASP13 group for the same target. For six of 13 targets, the most accurate model provided by any NMR-assisted prediction group was more accurate than the most accurate model provided by any regular prediction group; however, for the remaining seven targets, one or more regular prediction method provided a more accurate model than even the best NMR-assisted model. These results suggest a novel approach for protein structure determination, in which advanced prediction methods are first used to generate structural models, and sparse NMR data is then used to validate and/or refine these models.     

蛋白质结构预测的理论方法及阶段

一直以来,蛋白质结构预测都是人们研究的焦点,综述了蛋白质结构预测的几种理论方法和不同阶段。     

Protein tertiary structure modeling driven by deep learning and contact distance prediction in CASP13

Predicting residue-residue distance relationships (eg, contacts) has become the key direction to advance protein structure prediction since 2014 CASP11 experiment, while deep learning has revolutionized the technology for contact and distance distribution prediction since its debut in 2012 CASP10 experiment. During 2018 CASP13 experiment, we enhanced our MULTICOM protein structure prediction system with three major components: contact distance prediction based on deep convolutional neural networks, distance-driven template-free (ab initio) modeling, and protein model ranking empowered by deep learning and contact prediction. Our experiment demonstrates that contact distance prediction and deep learning methods are the key reasons that MULTICOM was ranked 3rd out of all 98 predictors in both template-free and template-based structure modeling in CASP13. Deep convolutional neural network can utilize global information in pairwise residue-residue features such as coevolution scores to substantially improve contact distance prediction, which played a decisive role in correctly folding some free modeling and hard template-based modeling targets. Deep learning also successfully integrated one-dimensional structural features, two-dimensional contact information, and three-dimensional structural quality scores to improve protein model quality assessment, where the contact prediction was demonstrated to consistently enhance ranking of protein models for the first time. The success of MULTICOM system clearly shows that protein contact distance prediction and model selection driven by deep learning holds the key of solving protein structure prediction problem. However, there are still challenges in accurately predicting protein contact distance when there are few homologous sequences, folding proteins from noisy contact distances, and ranking models of hard targets.     

Perfect temperature for protein structure prediction and folding

We have investigated the influence of the “noise” of inevitable errors in energetic parameters on-protein structure prediction. Because of this noise, only a part of all the interactions operating in a protein chain can be taken into account, and therefore a search for the energy minimum becomes inadequate for protein structure prediction. One can rather rely on statistical mechanics: a calculation carried out at a temperature T_* somewhat below that of protein melting gives the best possible, though always approximate prediction. The early stages of protein folding also “take into account” only a part of all the interactions; consequently, the same temperature T_* is favorable for the self-organization of native-like intermediates in protein folding. © 1995 Wiley-Liss, Inc.     

LiveBench-8: the large-scale, continuous assessment of automated protein structure prediction

We present the results of the evaluation of the latest LiveBench-8 experiment. These results provide a snapshot view of the state of the art in automated protein structure prediction, just before the 2004 CAFASP-4/CASP-6 experiments begin. The last CAFASP/CASP experiments demonstrated that automated meta-predictors entail a significant advance in the field, already challenging most human expert predictors. LiveBench-8 corroborates the superior performance of meta-predictors, which are able to produce useful predictions for over one-half of the test targets. More importantly, LiveBench-8 identifies a handful of recently developed autonomous (nonmeta) servers that perform at the very top, suggesting that further progress in the individual methods has recently been obtained.     

SimFold energy function for de novo protein structure prediction: consensus with Rosetta

Predicting protein tertiary structures by in silico folding is still very difficult for proteins that have new folds. Here, we developed a coarse-grained energy function, SimFold, for de novo structure prediction, performed a benchmark test of prediction with fragment assembly simulations for 38 test proteins, and proposed consensus prediction with Rosetta. The SimFold energy consists of many terms that take into account solvent-induced effects on the basis of physicochemical consideration. In the benchmark test, SimFold succeeded in predicting native structures within 6.5 A for 12 of 38 proteins; this success rate was the same as that by the publicly available version of Rosetta (ab initio version 1.2) run with default parameters. We investigated which energy terms in SimFold contribute to structure prediction performance, finding that the hydrophobic interaction is the most crucial for the prediction, whereas other sequence-specific terms have weak but positive roles. In the benchmark, well-predicted proteins by SimFold and by Rosetta were not the same for 5 of 12 proteins, which led us to introduce consensus prediction. With combined decoys, we succeeded in prediction for 16 proteins, four more than SimFold or Rosetta separately. For each of 38 proteins, structural ensembles generated by SimFold and by Rosetta were qualitatively compared by mapping sampled structural space onto two dimensions. For proteins of which one of the two methods succeeded and the other failed in prediction, the former had a less scattered ensemble located around the native. For proteins of which both methods succeeded in prediction, often two ensembles were mixed up.     

Constructing templates for protein structure prediction by simulation of protein folding pathways

How a one-dimensional protein sequence folds into a specific 3D structure remains a difficult challenge in structural biology. Many computational methods have been developed in an attempt to predict the tertiary structure of the protein; most of these employ approaches that are based on the accumulated knowledge of solved protein structures. Here we introduce a novel and fully automated approach for predicting the 3D structure of a protein that is based on the well accepted notion that protein folding is a hierarchical process. Our algorithm follows the hierarchical model by employing two stages: the first aims to find a match between the sequences of short independently-folding structural entities and parts of the target sequence and assigns the respective structures. The second assembles these local structural parts into a complete 3D structure, allowing for long-range interactions between them. We present the results of applying our method to a subset of the targets from CASP6 and CASP7. Our results indicate that for targets with a significant sequence similarity to known structures we are often able to provide predictions that are better than those achieved by two leading servers, and that the most significant improvements in comparison with these methods occur in regions of a gapped structural alignment between the native structure and the closest available structural template. We conclude that in addition to performing well for targets with known homologous structures, our method shows great promise for addressing the more general category of comparative modeling targets, which is our next goal.     

Contact order and ab initio protein structure prediction

Although much of the motivation for experimental studies of protein folding is to obtain insights for improving protein structure prediction, there has been relatively little connection between experimental protein folding studies and computational structural prediction work in recent years. In the present study, we show that the relationship between protein folding rates and the contact order (CO) of the native structure has implications for ab initio protein structure prediction. Rosetta ab initio folding simulations produce a dearth of high CO structures and an excess of low CO structures, as expected if the computer simulations mimic to some extent the actual folding process. Consistent with this, the majority of failures in ab initio prediction in the CASP4 (critical assessment of structure prediction) experiment involved high CO structures likely to fold much more slowly than the lower CO structures for which reasonable predictions were made. This bias against high CO structures can be partially alleviated by performing large numbers of additional simulations, selecting out the higher CO structures, and eliminating the very low CO structures; this leads to a modest improvement in prediction quality. More significant improvements in predictions for proteins with complex topologies may be possible following significant increases in high-performance computing power, which will be required for thoroughly sampling high CO conformations (high CO proteins can take six orders of magnitude longer to fold than low CO proteins). Importantly for such a strategy, simulations performed for high CO structures converge much less strongly than those for low CO structures, and hence, lack of simulation convergence can indicate the need for improved sampling of high CO conformations. The parallels between Rosetta simulations and folding in vivo may extend to misfolding: The very low CO structures that accumulate in Rosetta simulations consist primarily of local up-down beta-sheets that may resemble precursors to amyloid formation.     

Assessment of protein model structure accuracy estimation in CASP13: Challenges in the era of deep learning

Scoring model structure is an essential component of protein structure prediction that can affect the prediction accuracy tremendously. Users of protein structure prediction results also need to score models to select the best models for their application studies. In Critical Assessment of techniques for protein Structure Prediction (CASP), model accuracy estimation methods have been tested in a blind fashion by providing models submitted by the tertiary structure prediction servers for scoring. In CASP13, model accuracy estimation results were evaluated in terms of both global and local structure accuracy. Global structure accuracy estimation was evaluated by the quality of the models selected by the global structure scores and by the absolute estimates of the global scores. Residue-wise, local structure accuracy estimations were evaluated by three different measures. A new measure introduced in CASP13 evaluates the ability to predict inaccurately modeled regions that may be improved by refinement. An intensive comparative analysis on CASP13 and the previous CASPs revealed that the tertiary structure models generated by the CASP13 servers show very distinct features. Higher consensus toward models of higher global accuracy appeared even for free modeling targets, and many models of high global accuracy were not well optimized at the atomic level. This is related to the new technology in CASP13, deep learning for tertiary contact prediction. The tertiary model structures generated by deep learning pose a new challenge for EMA (estimation of model accuracy) method developers. Model accuracy estimation itself is also an area where deep learning can potentially have an impact, although current EMA methods have not fully explored that direction.     

Fragment-HMM: a new approach to protein structure prediction

We designed a simple position-specific hidden Markov model to predict protein structure. Our new framework naturally repeats itself to converge to a final target, conglomerating fragment assembly, clustering, target selection, refinement, and consensus, all in one process. Our initial implementation of this theory converges to within 6 A of the native structures for 100% of decoys on all six standard benchmark proteins used in ROSETTA (discussed by Simons and colleagues in a recent paper), which achieved only 14%-94% for the same data. The qualities of the best decoys and the final decoys our theory converges to are also notably better.