Dataset Information

The whole is greater than its parts: ensembling improves protein contact prediction.

ABSTRACT: The prediction of amino acid contacts from protein sequence is an important problem, as protein contacts are a vital step towards the prediction of folded protein structures. We propose that a powerful concept from deep learning, called ensembling, can increase the accuracy of protein contact predictions by combining the outputs of different neural network models. We show that ensembling the predictions made by different groups at the recent Critical Assessment of Protein Structure Prediction (CASP13) outperforms all individual groups. Further, we show that contacts derived from the distance predictions of three additional deep neural networks-AlphaFold, trRosetta, and ProSPr-can be substantially improved by ensembling all three networks. We also show that ensembling these recent deep neural networks with the best CASP13 group creates a superior contact prediction tool. Finally, we demonstrate that two ensembled networks can successfully differentiate between the folds of two highly homologous sequences. In order to build further on these findings, we propose the creation of a better protein contact benchmark set and additional open-source contact prediction methods.

SUBMITTER: Billings WM

PROVIDER: S-EPMC8044223 | biostudies-literature | 2021 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

The whole is greater than its parts: ensembling improves protein contact prediction.

Billings Wendy M WM Morris Connor J CJ Della Corte Dennis D

Scientific reports 20210413 1

The prediction of amino acid contacts from protein sequence is an important problem, as protein contacts are a vital step towards the prediction of folded protein structures. We propose that a powerful concept from deep learning, called ensembling, can increase the accuracy of protein contact predictions by combining the outputs of different neural network models. We show that ensembling the predictions made by different groups at the recent Critical Assessment of Protein Structure Prediction (C ...[more]

PMID: 33850214

Similar Datasets

Project description:BackgroundReconstructability of population history, from genetic information of extant individuals, is studied under a simulation setting. We do not address the issue of accuracy of the reconstruction algorithms: we assume the availability of the theoretical best algorithm. On the other hand, we focus on the fraction (1 - f) of the common genetic history that is irreconstructible or impenetrable. Thus the fraction, f, gives an upper bound on the extent of estimability. In other words, there exists no method that can reconstruct a fraction larger than f of the entire common genetic history. For the realization of such a study, we first define a natural measure of the amount of genetic history. Next, we use a population simulator (from literature) that has at least two features. Firstly, it has the capability of providing samples from different demographies, to effectively reflect reality. Secondly, it also provides the underlying relevant genetic history, captured in its entirety, where such a measure is applicable. Finally, to compute f, we use an information content measure of the relevant genetic history. The simulator of choice provided the following demographies: Africans, Europeans, Asians and Afro-Americans.ResultsWe observe that higher the rate of recombination, lower the value of f, while f is invariant over varying mutation rates, in each of the demographies. The value of f increases with the number of samples, reaching a plateau and suggesting that in all the demographies at least about one-third of the relevant genetic history is impenetrable. The most surprising observation is that the the sum of the reconstructible history of the subsegments is indeed larger than the reconstructible history of the whole segment. In particular, longer the chromosomal segment, smaller the value of f, in all the demographies.ConclusionsWe present the very first framework for measuring the fraction of the relevant genetic history of a population that is mathematically elusive. Our observed results on the tested demographies suggest that it may be better to aggregate the analysis of smaller chunks of chromosomal segments than fewer large chunks. Also, no matter the richness of samples in a population, at least one-third of the population genetic history is impenetrable. The framework also opens up possible new lines of investigation along the following. Given the characteristics of a population, possibly derived from observed extant individuals, to estimate the (1) optimal sample size and (2) optimal sequence length for the most informative analysis.

Project description:MotivationPrediction of protein-protein interaction has become an important part of systems biology in reverse engineering the biological networks for better understanding the molecular biology of the cell. Although significant progress has been made in terms of prediction accuracy, most computational methods only predict whether two proteins interact but not their interacting residues-the information that can be very valuable for understanding the interaction mechanisms and designing modulation of the interaction. In this work, we developed a computational method to predict the interacting residue pairs-contact matrix for interacting protein domains, whose rows and columns correspond to the residues in the two interacting domains respectively and whose values (1 or 0) indicate whether the corresponding residues (do or do not) interact.ResultsOur method is based on supervised learning using support vector machines. For each domain involved in a given domain-domain interaction (DDI), an interaction profile hidden Markov model (ipHMM) is first built for the domain family, and then each residue position for a member domain sequence is represented as a 20-dimension vector of Fisher scores, characterizing how similar it is as compared with the family profile at that position. Each element of the contact matrix for a sequence pair is now represented by a feature vector from concatenating the vectors of the two corresponding residues, and the task is to predict the element value (1 or 0) from the feature vector. A support vector machine is trained for a given DDI, using either a consensus contact matrix or contact matrices for individual sequence pairs, and is tested by leave-one-out cross validation. The performance averaged over a set of 115 DDIs collected from the 3 DID database shows significant improvement (sensitivity up to 85%, and specificity up to 85%), as compared with a multiple sequence alignment-based method (sensitivity 57%, and specificity 78%) previously reported in the literature.Supplementary informationSupplementary data are available at Bioinformatics online.

Dataset Information

The whole is greater than its parts: ensembling improves protein contact prediction.

Publications

The whole is greater than its parts: ensembling improves protein contact prediction.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets