DeepMSA: constructing deep multiple sequence alignment to improve contact prediction and fold-recognition for distant-homology proteins

By Chengxin Zhang, Wei Zheng, S. Mortuza, Yang Li, Yang Zhang. 2019

159 citations recorded by the source corpus.

MOTIVATION The success of genome sequencing techniques has resulted in rapid explosion of protein sequences. Collections of multiple homologous sequences can provide critical information to the modeling of structure and function of unknown proteins. There are however no standard and efficient pipeline available for sensitive multiple sequence alignment (MSA) collection. This is particularly challenging when large whole-genome and metagenome databases are involved. RESULTS We developed DeepMSA, a new open-source method for sensitive MSA construction, which has homologous sequences and alignments created from multi-sources of whole-genome and metagenome databases through complementary hidden Markov model (HMM) algorithms. The practical usefulness of the pipeline was examined in three large-scale benchmark experiments based on 614 non-redundant proteins. First, DeepMSA was utilized to generate MSAs for residue-level contact prediction by six coevolution and deep learning-based programs, which resulted in an accuracy increase in long-range contacts by up to 24.4% compared to the default programs. Next, multiple threading programs are performed for homologous structure identification, where the average TM-score of the template alignments has over 7.5% increases with the use of the new DeepMSA profiles. Finally, DeepMSA was used for secondary structure prediction and resulted in statistically significant improvements in the Q3 accuracy. It is noted that all these improvements were achieved without re-training the parameters and neural-network models, demonstrating the robustness and general usefulness of the DeepMSA in protein structural bioinformatics applications, especially for targets without homologous templates in the PDB library. AVAILABILITY https://zhanglab.ccmb.med.umich.edu/DeepMSA/. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

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

Citations in the corpus

CopulaNet: Learning residue co-evolution directly from multiple sequence alignment for protein structure prediction (2020, methodology): “For a target protein, we first search its homologue proteins by running DeepMSA [26] (with default parameters) against sequence databases uniclust30 (version 2017-10), uniref90 (version 2018-03) and metaclust50 (version 2018-01).”
Transformer protein language models are unsupervised structure learners (2020, methodology): “Additionally, we evaluated alternate MSAs by running Jackhmmer until a Neff of 128 was achieved (with a maximum of 8 iterations), a procedure described by Zhang et al. (2020).”
Integrating end-to-end learning with deep geometrical potentials for ab initio RNA structure prediction (2022, unknown): “…can bene ﬁ t from co-evolution information derived from MSAs and therefore often achieve better performance on overall structure prediction 18,38 , training on single query sequences has advantages in terms of the speed and ﬂ exibility of modeling as the procedure does not rely on the…”
Toward the solution of the protein structure prediction problem (2021, background): “Later MSA construction methods (183, 184) con ﬁ rmed the usefulness of metagenome-derived MSAs for improving contact prediction (168, 183, 184), threading results for distantly homologous targets (99, 184), and the ability to model proteins that belong to families with unknown structures (25, 185).”
Deep learning techniques have significantly impacted protein structure prediction and protein design. (2021, methodology): “…there was nearly no correlation (with a PCC = 0.145) between the TM-score of the AlphaFold2 models and the logarithm of the Neff (the number of effective sequences) of the multiple sequence alignments collected by the DeepMSA program [39] searched through metagenome sequence databases (Figure 2b).”
Folding non-homologous proteins by coupling deep-learning contact maps with I-TASSER assembly simulations (2021, methodology): “In Figure S2C, we present the contact accuracy results of the six contact predictors versus the number of effective sequences in the MSAs (or Neff, deﬁned in Equation S12 in ‘‘DeepMSA for MSA generation’’ under STAR Methods).”
Protein language model embeddings for fast, accurate, alignment-free protein structure prediction (2021, background): “solution is to generate multiple MSAs with different parameters, alignment tools and databases (Jain et al., 2021; Zhang et al., 2020), rendering the input generation even more time-consuming, e.”
I-TASSER-MTD: a deep-learning-based platform for multi-domain protein structure and function prediction (2022, methodology): “To create the deep-learning model, the multiple sequence alignment (MSA) is ﬁ rst created by iteratively searching the query against four metagenome sequence databases (Metaclust 48 , BFD 49 , Mgnify 50 and IMG/M 51 ) and two whole-genome databases (Uni-clust30 52 and Uniref90 53 ) using DeepMSA 54 ”
Ensembling multiple raw coevolutionary features with deep residual neural networks for contact-map prediction in CASP13 (2019, methodology): “27 First, HHblits 28 is used to search against UniClust30 29 for three iterations with a minimal coverage equal to 50%.”
Deducing high-accuracy protein contact-maps from a triplet of coevolutionary matrices through deep residual convolutional networks (2020, methodology): “The full 3-stage DeepMSA pipeline, on average, takes 1.32 hours [30] on its benchmark set.”