Improved Contact Predictions Using the Recognition of Protein Like Contact Patterns

By Marcin J. Skwark, Daniele Raimondi, Mirco Michel, Arne Elofsson. 2014

163 citations recorded by the source corpus.

Given sufficient large protein families, and using a global statistical inference approach, it is possible to obtain sufficient accuracy in protein residue contact predictions to predict the structure of many proteins. However, these approaches do not consider the fact that the contacts in a protein are neither randomly, nor independently distributed, but actually follow precise rules governed by the structure of the protein and thus are interdependent. Here, we present PconsC2, a novel method that uses a deep learning approach to identify protein-like contact patterns to improve contact predictions. A substantial enhancement can be seen for all contacts independently on the number of aligned sequences, residue separation or secondary structure type, but is largest for β-sheet containing proteins. In addition to being superior to earlier methods based on statistical inferences, in comparison to state of the art methods using machine learning, PconsC2 is superior for families with more than 100 effective sequence homologs. The improved contact prediction enables improved structure prediction.

Subfields: Materials Chemistry, Molecular 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 Materials Chemistry

In Molecular Biology

In Computational Biology & Bioinformatics

Citations in the corpus

Opportunities and obstacles for deep learning in biology and medicine (2017, methodology): “In recent years, deeper architectures have been explored for contact prediction, such as CMAPpro [196], DNCON [197] and PConsC [198].”
High-resolution de novo structure prediction from primary sequence (2022, background): “…S1 to S4 Supplementary Materials for High-resolution de novo protein structure prediction from primary sequence Unsupervised pre-trained models have achieved tremendous success in research areas such as natural language processing ( 1–4 ), computer vision ( 5–8 ) and single cell analysis ( 9, 10 ).”
Folding membrane proteins by deep transfer learning (2017, background): “…co-evolved residues, such as EVfold (Marks et al., 2011), PSICOV (Jones et al., 2012), CCMpred (Seemayer et al., 2014), Gremlin (Kamisetty et al., 2013) and others(Weigt et al., 2009), but it needs a large number of sequence homologs to be effective (Skwark et al., 2014, Ma et al., 2015).”
Accurate contact predictions using covariation techniques and machine learning (2015, background): “Therefore, like other authors,7,18,19 we calculate effective sequence counts ( N eff ) rather than raw sequence numbers to account for this redundancy.”
Limits and potential of combined folding and docking (2021, background): “Later, deep learning [4,5] and prediction of residue-residue distances provided further improvements [6,7].”
Inter-residue, inter-protein and inter-family coevolution: bridging the scales (2017, methodology): “contacts inside single proteins, an important improvement was obtained by supervised methods [46-48]: the results of DCA are put together with other features, which may include outcomes of other contact-prediction methods, predicted secondary structure or solvent accessibility.”
DESTINI: A deep-learning approach to contact-driven protein structure prediction (2019, methodology): “Moreover, template or sequence information can be combined to train supervised machine learning approaches, e.g., PconsC2 24 , MetaPSICOV 25 , SVMSEQ 11 , CMAPpro 26 , PhyCMAP 27 and CoinDCA-NN 28 .”
PconsC4: fast, free, easy, and accurate contact predictions (2018, methodology): “This problem has been overcome by reﬁning the initial DCA prediction with deep learning methods (Skwark et al. , 2014; Wang et al. , 2017).”
Large-scale structure prediction by improved contact predictions and model quality assessment (2017, methodology): “Until recently such methods have been limited to very large protein families (Kamisetty et al., 2013; Skwark et al., 2014).”
All-atom 3D structure prediction of transmembrane β-barrel proteins from sequences (2015, background): “There has also recently been some progress toward decreasing the need for the number of sequences by combining deep learning and direct information (41).”