skip to main content


Title: The choice of sequence homologs included in multiple sequence alignments has a dramatic impact on evolutionary conservation analysis
Abstract Motivation

The analysis of sequence conservation patterns has been widely utilized to identify functionally important (catalytic and ligand-binding) protein residues for over a half-century. Despite decades of development, on average state-of-the-art non-template-based functional residue prediction methods must predict ∼25% of a protein’s total residues to correctly identify half of the protein’s functional site residues. The overwhelming proportion of false positives results in reported ‘F-Scores’ of ∼0.3. We investigated the limits of current approaches, focusing on the so-far neglected impact of the specific choice of homologs included in multiple sequence alignments (MSAs).

Results

The limits of conservation-based functional residue prediction were explored by surveying the binding sites of 1023 proteins. A straightforward conservation analysis of MSAs composed of randomly selected homologs sampled from a PSI-BLAST search achieves average F-Scores of ∼0.3, a performance matching that reported by state-of-the-art methods, which often consider additional features for the prediction in a machine learning setting. Interestingly, we found that a simple combinatorial MSA sampling algorithm will in almost every case produce an MSA with an optimal set of homologs whose conservation analysis reaches average F-Scores of ∼0.6, doubling state-of-the-art performance. We also show that this is nearly at the theoretical limit of possible performance given the agreement between different binding site definitions. Additionally, we showcase the progress in this direction made by Selection of Alignment by Maximal Mutual Information (SAMMI), an information-theory-based approach to identifying biologically informative MSAs. This work highlights the importance and the unused potential of optimally composed MSAs for conservation analysis.

Supplementary information

Supplementary data are available at Bioinformatics online.

 
more » « less
NSF-PAR ID:
10393456
Author(s) / Creator(s):
; ;
Publisher / Repository:
Oxford University Press
Date Published:
Journal Name:
Bioinformatics
Volume:
35
Issue:
1
ISSN:
1367-4803
Page Range / eLocation ID:
p. 12-19
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract Motivation

    Multiple sequence alignments (MSAs) of homologous sequences contain information on structural and functional constraints and their evolutionary histories. Despite their importance for many downstream tasks, such as structure prediction, MSA generation is often treated as a separate pre-processing step, without any guidance from the application it will be used for.

    Results

    Here, we implement a smooth and differentiable version of the Smith–Waterman pairwise alignment algorithm that enables jointly learning an MSA and a downstream machine learning system in an end-to-end fashion. To demonstrate its utility, we introduce SMURF (Smooth Markov Unaligned Random Field), a new method that jointly learns an alignment and the parameters of a Markov Random Field for unsupervised contact prediction. We find that SMURF learns MSAs that mildly improve contact prediction on a diverse set of protein and RNA families. As a proof of concept, we demonstrate that by connecting our differentiable alignment module to AlphaFold2 and maximizing predicted confidence, we can learn MSAs that improve structure predictions over the initial MSAs. Interestingly, the alignments that improve AlphaFold predictions are self-inconsistent and can be viewed as adversarial. This work highlights the potential of differentiable dynamic programming to improve neural network pipelines that rely on an alignment and the potential dangers of optimizing predictions of protein sequences with methods that are not fully understood.

    Availability and implementation

    Our code and examples are available at: https://github.com/spetti/SMURF.

    Supplementary information

    Supplementary data are available at Bioinformatics online.

     
    more » « less
  2. Abstract Motivation

    Protein structure prediction has been greatly improved by deep learning, but the contribution of different information is yet to be fully understood. This article studies the impacts of two kinds of information for structure prediction: template and multiple sequence alignment (MSA) embedding. Templates have been used by some methods before, such as AlphaFold2, RoseTTAFold and RaptorX. AlphaFold2 and RosetTTAFold only used templates detected by HHsearch, which may not perform very well on some targets. In addition, sequence embedding generated by pre-trained protein language models has not been fully explored for structure prediction. In this article, we study the impact of templates (including the number of templates, the template quality and how the templates are generated) on protein structure prediction accuracy, especially when the templates are detected by methods other than HHsearch. We also study the impact of sequence embedding (generated by MSATransformer and ESM-1b) on structure prediction.

    Results

    We have implemented a deep learning method for protein structure prediction that may take templates and MSA embedding as extra inputs. We study the contribution of templates and MSA embedding to structure prediction accuracy. Our experimental results show that templates can improve structure prediction on 71 of 110 CASP13 (13th Critical Assessment of Structure Prediction) targets and 47 of 91 CASP14 targets, and templates are particularly useful for targets with similar templates. MSA embedding can improve structure prediction on 63 of 91 CASP14 (14th Critical Assessment of Structure Prediction) targets and 87 of 183 CAMEO targets and is particularly useful for proteins with shallow MSAs. When both templates and MSA embedding are used, our method can predict correct folds (TMscore > 0.5) for 16 of 23 CASP14 FM targets and 14 of 18 Continuous Automated Model Evaluation (CAMEO) targets, outperforming RoseTTAFold by 5% and 7%, respectively.

    Availability and implementation

    Available at https://github.com/xluo233/RaptorXFold.

    Supplementary information

    Supplementary data are available at Bioinformatics online.

     
    more » « less
  3. null (Ed.)
    Abstract Protein 3D structure prediction has advanced significantly in recent years due to improving contact prediction accuracy. This improvement has been largely due to deep learning approaches that predict inter-residue contacts and, more recently, distances using multiple sequence alignments (MSAs). In this work we present AttentiveDist, a novel approach that uses different MSAs generated with different E-values in a single model to increase the co-evolutionary information provided to the model. To determine the importance of each MSA’s feature at the inter-residue level, we added an attention layer to the deep neural network. We show that combining four MSAs of different E-value cutoffs improved the model prediction performance as compared to single E-value MSA features. A further improvement was observed when an attention layer was used and even more when additional prediction tasks of bond angle predictions were added. The improvement of distance predictions were successfully transferred to achieve better protein tertiary structure modeling. 
    more » « less
  4. Ponty, Yann (Ed.)
    Abstract Motivation Detecting subtle biologically relevant patterns in protein sequences often requires the construction of a large and accurate multiple sequence alignment (MSA). Methods for constructing MSAs are usually evaluated using benchmark alignments, which, however, typically contain very few sequences and are therefore inappropriate when dealing with large numbers of proteins. Results eCOMPASS addresses this problem using a statistical measure of relative alignment quality based on direct coupling analysis (DCA): To maintain protein structural integrity over evolutionary time, substitutions at one residue position typically result in compensating substitutions at other positions. eCOMPASS computes the statistical significance of the congruence between high scoring directly coupled pairs and 3D contacts in corresponding structures, which depends upon properly aligned homologous residues. We illustrate eCOMPASS using both simulated and real MSAs. Availability and Implementation The eCOMPASS executable, C ++ open source code and input data sets are available at https://www.igs.umaryland.edu/labs/neuwald/software/compass. Supplementary information Supplementary data are available at Bioinformatics online. 
    more » « less
  5. Abstract Summary

    A new dynamic community identifier (DCI) is presented that relies upon protein residue dynamic cross-correlations generated by Gaussian elastic network models to identify those residue clusters exhibiting motions within a protein. A number of examples of communities are shown for diverse proteins, including GPCRs. It is a tool that can immediately simplify and clarify the most essential functional moving parts of any given protein. Proteins usually can be subdivided into groups of residues that move as communities. These are usually densely packed local sub-structures, but in some cases can be physically distant residues identified to be within the same community. The set of these communities for each protein are the moving parts. The ways in which these are organized overall can aid in understanding many aspects of functional dynamics and allostery. DCI enables a more direct understanding of functions including enzyme activity, action across membranes and changes in the community structure from mutations or ligand binding. The DCI server is freely available on a web site (https://dci.bb.iastate.edu/).

    Supplementary information

    Supplementary data are available at Bioinformatics online.

     
    more » « less