Deep learning extends de novo protein modelling coverage of genomes using iteratively predicted structural constraints

By Joe G. Greener, Shaun M. Kandathil, David T. Jones. 2018

168 citations recorded by the source corpus.

The inapplicability of amino acid covariation methods to small protein families has limited their use for structural annotation of whole genomes. Recently, deep learning has shown promise in allowing accurate residue-residue contact prediction even for shallow sequence alignments. Here we introduce DMPfold, which uses deep learning to predict inter-atomic distance bounds, the main chain hydrogen bond network, and torsion angles, which it uses to build models in an iterative fashion. DMPfold produces more accurate models than two popular methods for a test set of CASP12 domains, and works just as well for transmembrane proteins. Applied to all Pfam domains without known structures, confident models for 25% of these so-called dark families were produced in under a week on a small 200 core cluster. DMPfold provides models for 16% of human proteome UniProt entries without structures, generates accurate models with fewer than 100 sequences in some cases, and is freely available. Prediction of protein structures on the scale of genomes remains a challenge. Here the authors introduce a protein structure prediction method that uses deep learning to predict inter-atomic distances, torsion angles and hydrogen bonds, and apply it to predict the structures of 1475 Pfam domains.

Subfields: Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

1. Yang Zhang
2. Robin Pearce
3. David T. Jones
4. Chengxin Zhang
5. Yang Li
6. Eric W. Bell
7. Joe G. Greener
8. Wei Zheng

Citations in the corpus

• Protein structure prediction using multiple deep neural networks in the 13th Critical Assessment of Protein Structure Prediction (CASP13) (2019, background): “10,32 While in some cases, these distances are highly constrained by secondary structure or clear coevolutionary signals, in most cases, they will be uncertain.”
• Folding non-homologous proteins by coupling deep-learning contact maps with I-TASSER assembly simulations (2021, methodology): “DMPfold (Greener et al., 2019) attempted to fold 5,214 families for which HHsearch (Söding, 2005) was unable to detect homologous templates and reported 1,475 foldable models and 231 novel folds.”
• AI in health and medicine (2022, background): “AI has enabled recent advances in the area of biochemistry, improving understanding of the structure and behavior of biomol-ecules 42–45 .”
• Toward the solution of the protein structure prediction problem (2021, methodology): “Distance map prediction jumped to the forefront of the fi eld during the CASP13 experiment in 2018, when three predictors (RaptorX-Contact (43), DMPfold (172), and AlphaFold (173)), extended the use of deep ResNets for contact prediction to distance prediction.”
• Deducing high-accuracy protein contact-maps from a triplet of coevolutionary matrices through deep residual convolutional networks (2020, methodology): “In CASP13, in addition to TripletRes_CASP13 that used an ensemble of PLM, COV and PRE features, the individual raw PLM and COV features have also been utilized by AlphaFold [35] and DMP [36], respectively.”
• Efficient generative modeling of protein sequences using simple autoregressive models (2021, methodology): “ods, which currently revolutionize the field of proteinstructure prediction [25, 46, 58, 61].”
• Limits and potential of combined folding and docking (2021, background): “Today this means that for many, if not most, individual proteins, it is possible to accurately predict the structure of its folded domains [8].”
• Deep learning-based prediction of protein structure using learned representations of multiple sequence alignments (2020, methodology): “As with DMPfold1 [9] , we also trained “iterative” versions of the neural net model, with an extra input feature channel carrying pairwise Cα distances from an existing 3-”
• Ultrafast end-to-end protein structure prediction enables high-throughput exploration of uncharacterized proteins (2022, unknown): “We produced updated models for 5,193 Pfam families, using the same MSAs as used in the original DMPfold1 paper (9).”
• Recent developments in deep learning applied to protein structure prediction (2019, methodology): “These so-called residual NNs (ResNets) are becoming the standard architecture for training very deep CNNs, and have been used in the best-performing methods for contact and distance prediction in CASP, 9,10,41,42,47,48 including those by the Zhang and A7D groups in CASP13.”
• Predicting changes in protein thermodynamic stability upon point mutation with deep 3D convolutional neural networks (2020, background): “They have recently proven to perform well in residue contact prediction (57585 60) and protein tertiary structure prediction (61-64).”