Simple chained guide trees give high-quality protein multiple sequence alignments

learnMSA2: deep protein multiple alignments with large language and hidden Markov models

MOTIVATION

For the alignment of large numbers of protein sequences, tools are predominant that decide to align two residues using only simple prior knowledge, e.g. amino acid substitution matrices, and using only part of the available data. The accuracy of state-of-the-art programs declines with decreasing sequence identity and when increasingly large numbers of sequences are aligned. Recently, transformer-based deep-learning models started to harness the vast amount of protein sequence data, resulting in powerful pretrained language models with the main purpose of generating high-dimensional numerical representations, embeddings, for individual sites that agglomerate evolutionary, structural, and biophysical information.

RESULTS

We extend the traditional profile hidden Markov model so that it takes as inputs unaligned protein sequences and the corresponding embeddings. We fit the model with gradient descent using our existing differentiable hidden Markov layer. All sequences and their embeddings are jointly aligned to a model of the protein family. We report that our upgraded HMM-based aligner, learnMSA2, combined with the ProtT5-XL protein language model aligns on average almost 6% points more columns correctly than the best amino acid-based competitor and scales well with sequence number. The relative advantage of learnMSA2 over other programs tends to be greater when the sequence identity is lower and when the number of sequences is larger. Our results strengthen the evidence on the rich information contained in protein language models' embeddings and their potential downstream impact on the field of bioinformatics. Availability and implementation:  https://github.com/Gaius-Augustus/learnMSA, PyPI and Bioconda, evaluation: https://github.com/felbecker/snakeMSA.

Structure and electromechanical coupling of a voltage-gated Na+/H+ exchanger

Voltage-sensing domains control the activation of voltage-gated ion channels, with a few exceptions1. One such exception is the sperm-specific Na+/H+ exchanger SLC9C1, which is the only known transporter to be regulated by voltage-sensing domains2-5. After hyperpolarization of sperm flagella, SLC9C1 becomes active, causing pH alkalinization and CatSper Ca2+ channel activation, which drives chemotaxis2,6. SLC9C1 activation is further regulated by cAMP2,7, which is produced by soluble adenyl cyclase (sAC). SLC9C1 is therefore an essential component of the pH-sAC-cAMP signalling pathway in metazoa8,9, required for sperm motility and fertilization4. Despite its importance, the molecular basis of SLC9C1 voltage activation is unclear. Here we report cryo-electron microscopy (cryo-EM) structures of sea urchin SLC9C1 in detergent and nanodiscs. We show that the voltage-sensing domains are positioned in an unusual configuration, sandwiching each side of the SLC9C1 homodimer. The S4 segment is very long, 90 Å in length, and connects the voltage-sensing domains to the cytoplasmic cyclic-nucleotide-binding domains. The S4 segment is in the up configuration-the inactive state of SLC9C1. Consistently, although a negatively charged cavity is accessible for Na+ to bind to the ion-transporting domains of SLC9C1, an intracellular helix connected to S4 restricts their movement. On the basis of the differences in the cryo-EM structure of SLC9C1 in the presence of cAMP, we propose that, upon hyperpolarization, the S4 segment moves down, removing this constriction and enabling Na+/H+ exchange.

Auranofin targets UBA1 and enhances UBA1 activity by facilitating ubiquitin trans-thioesterification to E2 ubiquitin-conjugating enzymes

UBA1 is the primary E1 ubiquitin-activating enzyme responsible for generation of activated ubiquitin required for ubiquitination, a process that regulates stability and function of numerous proteins. Decreased or insufficient ubiquitination can cause or drive aging and many diseases. Therefore, a small-molecule enhancing UBA1 activity could have broad therapeutic potential. Here we report that auranofin, a drug approved for the treatment of rheumatoid arthritis, is a potent UBA1 activity enhancer. Auranofin binds to the UBA1's ubiquitin fold domain and conjugates to Cys1039 residue. The binding enhances UBA1 interactions with at least 20 different E2 ubiquitin-conjugating enzymes, facilitating ubiquitin charging to E2 and increasing the activities of seven representative E3s in vitro. Auranofin promotes ubiquitination and degradation of misfolded ER proteins during ER-associated degradation in cells at low nanomolar concentrations. It also facilitates outer mitochondrial membrane-associated degradation. These findings suggest that auranofin can serve as a much-needed tool for UBA1 research and therapeutic exploration.

Towards the accurate alignment of over a million protein sequences: Current state of the art

Large-scale genomics requires highly scalable and accurate multiple sequence alignment methods. Results collected over this last decade suggest accuracy loss when scaling up over a few thousand sequences. This issue has been actively addressed with a number of innovative algorithmic solutions that combine low-level hardware optimization with novel higher-level heuristics. This review provides an extensive critical overview of these recent methods. Using established reference datasets we conclude that albeit significant progress has been achieved, a unified framework able to consistently and efficiently produce high-accuracy large-scale multiple alignments is still lacking.

Developments in Algorithms for Sequence Alignment: A Review

The continuous development of sequencing technologies has enabled researchers to obtain large amounts of biological sequence data, and this has resulted in increasing demands for software that can perform sequence alignment fast and accurately. A number of algorithms and tools for sequence alignment have been designed to meet the various needs of biologists. Here, the ideas that prevail in the research of sequence alignment and some quality estimation methods for multiple sequence alignment tools are summarized.

ProPIP: a tool for progressive multiple sequence alignment with Poisson Indel Process

BACKGROUND

Current alignment tools typically lack an explicit model of indel evolution, leading to artificially short inferred alignments (i.e., over-alignment) due to inconsistencies between the indel history and the phylogeny relating the input sequences.

RESULTS

We present a new progressive multiple sequence alignment tool ProPIP. The process of insertions and deletions is described using an explicit evolutionary model-the Poisson Indel Process or PIP. The method is based on dynamic programming and is implemented in a frequentist framework. The source code can be compiled on Linux, macOS and Microsoft Windows platforms. The algorithm is implemented in C++ as standalone program. The source code is freely available on GitHub at https://github.com/acg-team/ProPIP and is distributed under the terms of the GNU GPL v3 license.

CONCLUSIONS

The use of an explicit indel evolution model allows to avoid over-alignment, to infer gaps in a phylogenetically consistent way and to make inferences about the rates of insertions and deletions. Instead of the arbitrary gap penalties, the parameters used by ProPIP are the insertion and deletion rates, which have biological interpretation and are contextualized in a probabilistic environment. As a result, indel rate settings may be optimised in order to infer phylogenetically meaningful gap patterns.

The Clustal Omega Multiple Alignment Package

Clustal Omega is a version, completely rewritten and revised in 2011, of the widely used Clustal series of programs for multiple sequence alignment. It can deal with very large numbers (many tens of thousands) of DNA/RNA or protein sequences due to its use of the mBed algorithm for calculating guide-trees. This algorithm allows very large alignment problems to be tackled very quickly, even on personal computers. The accuracy of the program has been considerably improved over earlier Clustal programs, through the use of the HHalign method for aligning profile hidden Markov models. The program currently is used from the command-line or can be run online.

MAFFT online service: multiple sequence alignment, interactive sequence choice and visualization

This article describes several features in the MAFFT online service for multiple sequence alignment (MSA). As a result of recent advances in sequencing technologies, huge numbers of biological sequences are available and the need for MSAs with large numbers of sequences is increasing. To extract biologically relevant information from such data, sophistication of algorithms is necessary but not sufficient. Intuitive and interactive tools for experimental biologists to semiautomatically handle large data are becoming important. We are working on development of MAFFT toward these two directions. Here, we explain (i) the Web interface for recently developed options for large data and (ii) interactive usage to refine sequence data sets and MSAs.

Evaluating Statistical Multiple Sequence Alignment in Comparison to Other Alignment Methods on Protein Data Sets

The estimation of multiple sequence alignments of protein sequences is a basic step in many bioinformatics pipelines, including protein structure prediction, protein family identification, and phylogeny estimation. Statistical coestimation of alignments and trees under stochastic models of sequence evolution has long been considered the most rigorous technique for estimating alignments and trees, but little is known about the accuracy of such methods on biological benchmarks. We report the results of an extensive study evaluating the most popular protein alignment methods as well as the statistical coestimation method BAli-Phy on 1192 protein data sets from established benchmarks as well as on 120 simulated data sets. Our study (which used more than 230 CPU years for the BAli-Phy analyses alone) shows that BAli-Phy has better precision and recall (with respect to the true alignments) than the other alignment methods on the simulated data sets but has consistently lower recall on the biological benchmarks (with respect to the reference alignments) than many of the other methods. In other words, we find that BAli-Phy systematically underaligns when operating on biological sequence data but shows no sign of this on simulated data. There are several potential causes for this change in performance, including model misspecification, errors in the reference alignments, and conflicts between structural alignment and evolutionary alignments, and future research is needed to determine the most likely explanation. We conclude with a discussion of the potential ramifications for each of these possibilities. [BAli-Phy; homology; multiple sequence alignment; protein sequences; structural alignment.]

Systematic benchmarking of omics computational tools

Computational omics methods packaged as software have become essential to modern biological research. The increasing dependence of scientists on these powerful software tools creates a need for systematic assessment of these methods, known as benchmarking. Adopting a standardized benchmarking practice could help researchers who use omics data to better leverage recent technological innovations. Our review summarizes benchmarking practices from 25 recent studies and discusses the challenges, advantages, and limitations of benchmarking across various domains of biology. We also propose principles that can make computational biology benchmarking studies more sustainable and reproducible, ultimately increasing the transparency of biomedical data and results.

Benchmarking studies are important for comprehensively understanding and evaluating different computational omics methods. Here, the authors review practices from 25 recent studies and propose principles to improve the quality of benchmarking studies.

Generalized Bootstrap Supports for Phylogenetic Analyses of Protein Sequences Incorporating Alignment Uncertainty

Phylogenetic reconstructions are essential in genomics data analyses and depend on accurate multiple sequence alignment (MSA) models. We show that all currently available large-scale progressive multiple alignment methods are numerically unstable when dealing with amino-acid sequences. They produce significantly different output when changing sequence input order. We used the HOMFAM protein sequences dataset to show that on datasets larger than 100 sequences, this instability affects on average 21.5% of the aligned residues. The resulting Maximum Likelihood (ML) trees estimated from these MSAs are equally unstable with over 38% of the branches being sensitive to the sequence input order. We established that about two-thirds of this uncertainty stems from the unordered nature of children nodes within the guide trees used to estimate MSAs. To quantify this uncertainty we developed unistrap, a novel approach that estimates the combined effect of alignment uncertainty and site sampling on phylogenetic tree branch supports. Compared with the regular bootstrap procedure, unistrap provides branch support estimates that take into account a larger fraction of the parameters impacting tree instability when processing datasets containing a large number of sequences.

Progressive multiple sequence alignment with indel evolution

Background

Sequence alignment is crucial in genomics studies. However, optimal multiple sequence alignment (MSA) is NP-hard. Thus, modern MSA methods employ progressive heuristics, breaking the problem into a series of pairwise alignments guided by a phylogeny. Changes between homologous characters are typically modelled by a Markov substitution model. In contrast, the dynamics of indels are not modelled explicitly, because the computation of the marginal likelihood under such models has exponential time complexity in the number of taxa. But the failure to model indel evolution may lead to artificially short alignments due to biased indel placement, inconsistent with phylogenetic relationship.

Results

Recently, the classical indel model TKF91 was modified to describe indel evolution on a phylogeny via a Poisson process, termed PIP. PIP allows to compute the joint marginal probability of an MSA and a tree in linear time. We present a new dynamic programming algorithm to align two MSAs –represented by the underlying homology paths– by full maximum likelihood under PIP in polynomial time, and apply it progressively along a guide tree. We have corroborated the correctness of our method by simulation, and compared it with competitive methods on an illustrative real dataset.

Conclusions

Our MSA method is the first polynomial time progressive aligner with a rigorous mathematical formulation of indel evolution. The new method infers phylogenetically meaningful gap patterns alternative to the popular PRANK, while producing alignments of similar length. Moreover, the inferred gap patterns agree with what was predicted qualitatively by previous studies. The algorithm is implemented in a standalone C++ program: https://github.com/acg-team/ProPIP. Supplementary data are available at BMC Bioinformatics online.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2357-1) contains supplementary material, which is available to authorized users.

Clustal Omega for making accurate alignments of many protein sequences

Clustal Omega is a widely used package for carrying out multiple sequence alignment. Here, we describe some recent additions to the package and benchmark some alternative ways of making alignments. These benchmarks are based on protein structure comparisons or predictions and include a recently described method based on secondary structure prediction. In general, Clustal Omega is fast enough to make very large alignments and the accuracy of protein alignments is high when compared to alternative packages. The package is freely available as executables or source code from www.clustal.org or can be run on-line from a variety of sites, especially the EBI www.ebi.ac.uk.

Computational complexity of algorithms for sequence comparison, short-read assembly and genome alignment

A multitude of algorithms for sequence comparison, short-read assembly and whole-genome alignment have been developed in the general context of molecular biology, to support technology development for high-throughput sequencing, numerous applications in genome biology and fundamental research on comparative genomics. The computational complexity of these algorithms has been previously reported in original research papers, yet this often neglected property has not been reviewed previously in a systematic manner and for a wider audience. We provide a review of space and time complexity of key sequence analysis algorithms and highlight their properties in a comprehensive manner, in order to identify potential opportunities for further research in algorithm or data structure optimization. The complexity aspect is poised to become pivotal as we will be facing challenges related to the continuous increase of genomic data on unprecedented scales and complexity in the foreseeable future, when robust biological simulation at the cell level and above becomes a reality.

QuickProbs 2: Towards rapid construction of high-quality alignments of large protein families

The ever-increasing size of sequence databases caused by the development of high throughput sequencing, poses to multiple alignment algorithms one of the greatest challenges yet. As we show, well-established techniques employed for increasing alignment quality, i.e., refinement and consistency, are ineffective when large protein families are investigated. We present QuickProbs 2, an algorithm for multiple sequence alignment. Based on probabilistic models, equipped with novel column-oriented refinement and selective consistency, it offers outstanding accuracy. When analysing hundreds of sequences, Quick-Probs 2 is noticeably better than ClustalΩ and MAFFT, the previous leaders for processing numerous protein families. In the case of smaller sets, for which consistency-based methods are the best performing, QuickProbs 2 is also superior to the competitors. Due to low computational requirements of selective consistency and utilization of massively parallel architectures, presented algorithm has similar execution times to ClustalΩ, and is orders of magnitude faster than full consistency approaches, like MSAProbs or PicXAA. All these make QuickProbs 2 an excellent tool for aligning families ranging from few, to hundreds of proteins.

FAMSA: Fast and accurate multiple sequence alignment of huge protein families

Rapid development of modern sequencing platforms has contributed to the unprecedented growth of protein families databases. The abundance of sets containing hundreds of thousands of sequences is a formidable challenge for multiple sequence alignment algorithms. The article introduces FAMSA, a new progressive algorithm designed for fast and accurate alignment of thousands of protein sequences. Its features include the utilization of the longest common subsequence measure for determining pairwise similarities, a novel method of evaluating gap costs, and a new iterative refinement scheme. What matters is that its implementation is highly optimized and parallelized to make the most of modern computer platforms. Thanks to the above, quality indicators, i.e. sum-of-pairs and total-column scores, show FAMSA to be superior to competing algorithms, such as Clustal Omega or MAFFT for datasets exceeding a few thousand sequences. Quality does not compromise on time or memory requirements, which are an order of magnitude lower than those in the existing solutions. For example, a family of 415519 sequences was analyzed in less than two hours and required no more than 8 GB of RAM. FAMSA is available for free at http://sun.aei.polsl.pl/REFRESH/famsa.

Application of the MAFFT sequence alignment program to large data-reexamination of the usefulness of chained guide trees

Motivation: Large multiple sequence alignments (MSAs), consisting of thousands of sequences, are becoming more and more common, due to advances in sequencing technologies. The MAFFT MSA program has several options for building large MSAs, but their performances have not been sufficiently assessed yet, because realistic benchmarking of large MSAs has been difficult. Recently, such assessments have been made possible through the HomFam and ContTest benchmark protein datasets. Along with the development of these datasets, an interesting theory was proposed: chained guide trees increase the accuracy of MSAs of structurally conserved regions. This theory challenges the basis of progressive alignment methods and needs to be examined by being compared with other known methods including computationally intensive ones.

Results: We used HomFam, ContTest and OXFam (an extended version of OXBench) to evaluate several methods enabled in MAFFT: (1) a progressive method with approximate guide trees, (2) a progressive method with chained guide trees, (3) a combination of an iterative refinement method and a progressive method and (4) a less approximate progressive method that uses a rigorous guide tree and consistency score. Other programs, Clustal Omega and UPP, available for large MSAs, were also included into the comparison. The effect of method 2 (chained guide trees) was positive in ContTest but negative in HomFam and OXFam. Methods 3 and 4 increased the benchmark scores more consistently than method 2 for the three datasets, suggesting that they are safer to use.

Availability and Implementation:http://mafft.cbrc.jp/alignment/software/

Contact:katoh@ifrec.osaka-u.ac.jp

Supplementary information:Supplementary data are available at Bioinformatics online.

Bayesian Top-Down Protein Sequence Alignment with Inferred Position-Specific Gap Penalties

We describe a Bayesian Markov chain Monte Carlo (MCMC) sampler for protein multiple sequence alignment (MSA) that, as implemented in the program GISMO and applied to large numbers of diverse sequences, is more accurate than the popular MSA programs MUSCLE, MAFFT, Clustal-Ω and Kalign. Features of GISMO central to its performance are: (i) It employs a "top-down" strategy with a favorable asymptotic time complexity that first identifies regions generally shared by all the input sequences, and then realigns closely related subgroups in tandem. (ii) It infers position-specific gap penalties that favor insertions or deletions (indels) within each sequence at alignment positions in which indels are invoked in other sequences. This favors the placement of insertions between conserved blocks, which can be understood as making up the proteins' structural core. (iii) It uses a Bayesian statistical measure of alignment quality based on the minimum description length principle and on Dirichlet mixture priors. Consequently, GISMO aligns sequence regions only when statistically justified. This is unlike methods based on the ad hoc, but widely used, sum-of-the-pairs scoring system, which will align random sequences. (iv) It defines a system for exploring alignment space that provides natural avenues for further experimentation through the development of new sampling strategies for more efficiently escaping from suboptimal traps. GISMO's superior performance is illustrated using 408 protein sets containing, on average, 235 sequences. These sets correspond to NCBI Conserved Domain Database alignments, which have been manually curated in the light of available crystal structures, and thus provide a means to assess alignment accuracy. GISMO fills a different niche than other MSA programs, namely identifying and aligning a conserved domain present within a large, diverse set of full length sequences. The GISMO program is available at http://gismo.igs.umaryland.edu/.

Using de novo protein structure predictions to measure the quality of very large multiple sequence alignments

Motivation: Multiple sequence alignments (MSAs) with large numbers of sequences are now commonplace. However, current multiple alignment benchmarks are ill-suited for testing these types of alignments, as test cases either contain a very small number of sequences or are based purely on simulation rather than empirical data.

Results: We take advantage of recent developments in protein structure prediction methods to create a benchmark (ContTest) for protein MSAs containing many thousands of sequences in each test case and which is based on empirical biological data. We rank popular MSA methods using this benchmark and verify a recent result showing that chained guide trees increase the accuracy of progressive alignment packages on datasets with thousands of proteins.

Availability and implementation: Benchmark data and scripts are available for download at http://www.bioinf.ucd.ie/download/ContTest.tar.gz.

Contact:des.higgins@ucd.ie

Supplementary information:Supplementary data are available at Bioinformatics online.

Instability in progressive multiple sequence alignment algorithms

BACKGROUND

Progressive alignment is the standard approach used to align large numbers of sequences. As with all heuristics, this involves a tradeoff between alignment accuracy and computation time.

RESULTS

We examine this tradeoff and find that, because of a loss of information in the early steps of the approach, the alignments generated by the most common multiple sequence alignment programs are inherently unstable, and simply reversing the order of the sequences in the input file will cause a different alignment to be generated. Although this effect is more obvious with larger numbers of sequences, it can also be seen with data sets in the order of one hundred sequences. We also outline the means to determine the number of sequences in a data set beyond which the probability of instability will become more pronounced.

CONCLUSIONS

This has major ramifications for both the designers of large-scale multiple sequence alignment algorithms, and for the users of these alignments.

DECIPHER: harnessing local sequence context to improve protein multiple sequence alignment

BACKGROUND

Alignment of large and diverse sequence sets is a common task in biological investigations, yet there remains considerable room for improvement in alignment quality. Multiple sequence alignment programs tend to reach maximal accuracy when aligning only a few sequences, and then diminish steadily as more sequences are added. This drop in accuracy can be partly attributed to a build-up of error and ambiguity as more sequences are aligned. Most high-throughput sequence alignment algorithms do not use contextual information under the assumption that sites are independent. This study examines the extent to which local sequence context can be exploited to improve the quality of large multiple sequence alignments.

RESULTS

Two predictors based on local sequence context were assessed: (i) single sequence secondary structure predictions, and (ii) modulation of gap costs according to the surrounding residues. The results indicate that context-based predictors have appreciable information content that can be utilized to create more accurate alignments. Furthermore, local context becomes more informative as the number of sequences increases, enabling more accurate protein alignments of large empirical benchmarks. These discoveries became the basis for DECIPHER, a new context-aware program for sequence alignment, which outperformed other programs on large sequence sets.

CONCLUSIONS

Predicting secondary structure based on local sequence context is an efficient means of breaking the independence assumption in alignment. Since secondary structure is more conserved than primary sequence, it can be leveraged to improve the alignment of distantly related proteins. Moreover, secondary structure predictions increase in accuracy as more sequences are used in the prediction. This enables the scalable generation of large sequence alignments that maintain high accuracy even on diverse sequence sets. The DECIPHER R package and source code are freely available for download at DECIPHER.cee.wisc.edu and from the Bioconductor repository.

Clustal omega

Clustal Omega is a package for making multiple sequence alignments of amino acid or nucleotide sequences, quickly and accurately. It is a complete upgrade and rewrite of earlier Clustal programs. This unit describes how to run Clustal Omega interactively from a command line, although it can also be run online from several sites. The unit describes a basic protocol for taking a set of unaligned sequences and producing a full alignment. There are also protocols for using an external HMM or iteration to help improve an alignment.

Systematic exploration of guide-tree topology effects for small protein alignments

Background

Guide-trees are used as part of an essential heuristic to enable the calculation of multiple sequence alignments. They have been the focus of much method development but there has been little effort at determining systematically, which guide-trees, if any, give the best alignments. Some guide-tree construction schemes are based on pair-wise distances amongst unaligned sequences. Others try to emulate an underlying evolutionary tree and involve various iteration methods.

Results

We explore all possible guide-trees for a set of protein alignments of up to eight sequences. We find that pairwise distance based default guide-trees sometimes outperform evolutionary guide-trees, as measured by structure derived reference alignments. However, default guide-trees fall way short of the optimum attainable scores. On average chained guide-trees perform better than balanced ones but are not better than default guide-trees for small alignments.

Conclusions

Alignment methods that use Consistency or hidden Markov models to make alignments are less susceptible to sub-optimal guide-trees than simpler methods, that basically use conventional sequence alignment between profiles. The latter appear to be affected positively by evolutionary based guide-trees for difficult alignments and negatively for easy alignments. One phylogeny aware alignment program can strongly discriminate between good and bad guide-trees. The results for randomly chained guide-trees improve with the number of sequences.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2105-15-338) contains supplementary material, which is available to authorized users.