High precision in protein contact prediction using fully convolutional neural networks and minimal sequence features

Jones, David T.; Kandathil, Shaun M.

doi:10.1093/bioinformatics/bty341

Cited by 177 publications

(198 citation statements)

References 45 publications

(57 reference statements)

Supporting

Mentioning

194

Contrasting

Unclassified

Order By: Relevance

“…Other features such as the outputs from PSICOV, CCMpred and FreeContact are used without modification, since they are defined on residue pairs. The 58‐channel MetaPSICOV inputs are combined with the 441‐channel DeepCov covariance matrices, which contain raw covariance values calculated for each pair of positions in the sequence alignment, for each pair of residue types . Two additional channels encode sequence separation between residue pairs and the sequence bounds; the latter is simply a channel where all input values are set to 1.…”

Section: Methodsmentioning

confidence: 99%

“…For our contact prediction effort in CASP13, we developed DeepMetaPSICOV (abbreviated DMP), a contact predictor based on a deep, fully convolutional residual network and a large input feature set. DMP is a logical extension and combination of our previous methods MetaPSICOV and DeepCov . The method is capable of precise predictions for a variety of proteins, including membrane proteins and those with relatively shallow sequence alignments.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Prediction of interresidue contacts with DeepMetaPSICOV in CASP13

2019

Self Cite

View full text Add to dashboard Cite

In this article, we describe our efforts in contact prediction in the CASP13 experiment. We employed a new deep learning‐based contact prediction tool, DeepMetaPSICOV (or DMP for short), together with new methods and data sources for alignment generation. DMP evolved from MetaPSICOV and DeepCov and combines the input feature sets used by these methods as input to a deep, fully convolutional residual neural network. We also improved our method for multiple sequence alignment generation and included metagenomic sequences in the search. We discuss successes and failures of our approach and identify areas where further improvements may be possible. DMP is freely available at: https://github.com/psipred/DeepMetaPSICOV.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Prediction of interresidue contacts with DeepMetaPSICOV in CASP13

2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…There are three coevolutionary features used by our methods. The first one, COV, is the covariance matrix as proposed by DeepCov . Considering an MSA with N rows and L columns, we can compute a 21∙ L by 21∙ L sample covariance matrix as follows:

S_{ij}^{ab} = f_{i, j} ((), a, b) - f_{i} ((), a) f_{j} ((), b)

where f i , j ( a , b ) is the observed relative frequency of residue pair a and b at position i and j and f i ( a ) is the frequency of occurrence of a residue type a at position i .…”

Section: Methodsmentioning

confidence: 99%

“…Unlike many other machine learning predictors, the recently developed DeepCov directly uses the raw sequence covariance matrix as its only feature, followed by convolutional neural networks to predict the contact‐map; it achieved comparable results to predictors based on features of post‐processed coevolutionary analysis. Alternatively, ResPRE considers the ridge estimation of the inverse of the covariance matrix, which was shown to be capable of wiping out noisy signal from translational interactions; when coupled with a fully residual neural network structure, the approach demonstrated superiority to the state‐of‐the‐art of other approaches.…”

Section: Introductionmentioning

confidence: 99%

Ensembling multiple raw coevolutionary features with deep residual neural networks for contact‐map prediction in CASP13

et al. 2019

View full text Add to dashboard Cite

We report the results of residue‐residue contact prediction of a new pipeline built purely on the learning of coevolutionary features in the CASP13 experiment. For a query sequence, the pipeline starts with the collection of multiple sequence alignments (MSAs) from multiple genome and metagenome sequence databases using two complementary Hidden Markov Model (HMM)‐based searching tools. Three profile matrices, built on covariance, precision, and pseudolikelihood maximization respectively, are then created from the MSAs, which are used as the input features of a deep residual convolutional neural network architecture for contact‐map training and prediction. Two ensembling strategies have been proposed to integrate the matrix features through end‐to‐end training and stacking, resulting in two complementary programs called TripletRes and ResTriplet, respectively. For the 31 free‐modeling domains that do not have homologous templates in the PDB, TripletRes and ResTriplet generated comparable results with an average accuracy of 0.640 and 0.646, respectively, for the top L/5 long‐range predictions, where 71% and 74% of the cases have an accuracy above 0.5. Detailed data analyses showed that the strength of the pipeline is due to the sensitive MSA construction and the advanced strategies for coevolutionary feature ensembling. Domain splitting was also found to help enhance the contact prediction performance. Nevertheless, contact models for tail regions, which often involve a high number of alignment gaps, and for targets with few homologous sequences are still suboptimal. Development of new approaches where the model is specifically trained on these regions and targets might help address these problems.

show abstract

“…One possible reason is that it did not use a metagenomics sequence database that contains sequences not present in the nonredundant protein sequence database and the latest HHblits database to collect homologous sequences. Another possible reason is the convolutional architecture used by DNCON2 is not deep enough in comparison with some other approaches . The second limitation is that only the coarse distance restraints derived from binary contacts at 8 Å threshold were used with CONFOLD2 for ab initio modeling, without taking advantage of the more detailed distance distribution spanning multiple distance thresholds predicted by DNCON2, which limited its capability to build quality models …”

Section: Resultsmentioning

confidence: 99%

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

et al. 2019

View full text Add to dashboard Cite

Predicting residue-residue distance relationships (e.g. 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 co-evolution 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 1D structural features, 2D contact information, and 3D 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.

show abstract

High precision in protein contact prediction using fully convolutional neural networks and minimal sequence features

Abstract: Supplementary data are available at Bioinformatics online.

Cited by 177 publications

References 45 publications

Prediction of interresidue contacts with DeepMetaPSICOV in CASP13

Prediction of interresidue contacts with DeepMetaPSICOV in CASP13

Ensembling multiple raw coevolutionary features with deep residual neural networks for contact‐map prediction in CASP13

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

Contact Info

Product

Resources

About