Prediction of interresidue contacts with DeepMetaPSICOV in CASP13

By Shaun M. Kandathil, Joe G. Greener, David T. Jones. 2019

99 citations recorded by the source corpus.

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.

Subfields: Molecular Biology, Structural Biology, Computational Biology & Bioinformatics

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. Andriy Kryshtafovych
4. Torsten Schwede
5. Krzysztof Fidelis
6. David T. Jones
7. John Moult
8. Jianlin Cheng

In Structural Biology

1. Yang Zhang
2. David T. Jones
3. Shaun M. Kandathil
4. Joe G. Greener
5. Robin Pearce
6. Maya Topf

In Computational Biology & Bioinformatics

1. Yang Zhang
2. Robin Pearce
3. Andriy Kryshtafovych
4. Torsten Schwede
5. David T. Jones
6. Krzysztof Fidelis
7. Joe G. Greener
8. Badri Adhikari

Citations in the corpus

• Improved protein structure prediction using predicted interresidue orientations (2019, neutral): “The overall architecture of the network is similar to those recently described for distance and contact prediction (3, 4, 7, 12).”
• Critical Assessment of Methods of Protein Structure Prediction (CASP) – Round XIV (2021, background): “In 2018 (CASP13), accuracy improved again, to around 70%, this time as the result of the use of deep-learning convolutional network methods.(35) A number of CASP participants also began using these methods to predict a continuous probability function for inter-residue distances rather than just a bi”
• Deep learning-based prediction of protein structure using learned representations of multiple sequence alignments (2020, methodology): “These embeddings are striped vertically and horizontally (similar to the manner in which per-residue features are prepared for use by 2D convolutional layers ​[4,5]​), and combined with a fast approximation of a precision matrix, calculated using the fast_dca algorithm from trRosetta [10]​.”
• Deep learning extends de novo protein modelling coverage of genomes using iteratively predicted structural constraints (2018, methodology): “DMPfold uses a set of deep neural networks similar in architecture and methods of training to our DeepMetaPSICOV (DMP) contact prediction method, which is described separately ​[17]​.”
• Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants (2021, methodology): “Box plots, line plots and bar plots were generated with the ggplot2 package 68 in R (r-project.org), and heatmaps were generated with the ComplexHeatmap package 69 in R. Pearson correlations and their p values were calculated with ggpubr package in R. Wilcoxon rank-sum test was performed using…”
• Toward the solution of the protein structure prediction problem (2021, background): “The demonstrated power of ResNets has inspired the vast majority of top ranked methods (165 – 167) developed since CASP12 to incorporate them into their architectures.”