Protein Sequence Space Research Articles

The structure of protein dynamic space

We use a bioinformatic description of amino acid dynamic properties, based on residue-specific average B factors, to construct a dynamics-based, large-scale description of a space of protein sequences. We examine the relationship between that space and an independently constructed, structure-based space comprising the same sequences. It is demonstrated that structure and dynamics are only moderately correlated. It is further shown that helical proteins fall into two classes with very different structure-dynamics relationships. We suggest that dynamics in the two helical classes are dominated by distinctly different modes--pseudo-one-dimensional, localized helical modes in one case, and pseudo-three-dimensional (3D) global modes in the other. Sheet/barrel and mixed-α/β proteins exhibit more conventional structure-dynamics relationships. It is found that the strongest correlation between structure and dynamic properties arises when the latter are represented by the sequence average of the dynamic index, which corresponds physically to the overall mobility of the protein. None of these results are accessible to bioinformatic methods hitherto available.

A Generative Neural Network for Maximizing Fitness and Diversity of Synthetic DNA and Protein Sequences.

SUMMARYEngineering gene and protein sequences with defined functional properties is a major goal of synthetic biology. Deep neural network models, together with gradient ascent-style optimization, show promise for sequence design. The generated sequences can however get stuck in local minima and often have low diversity. Here, we develop deep exploration networks (DENs), a class of activation-maximizing generative models, which minimize the cost of a neural network fitness predictor by gradient descent. By penalizing any two generated patterns on the basis of a similarity metric, DENs explicitly maximize sequence diversity. To avoid drifting into low-confidence regions of the predictor, we incorporate variational autoencoders to maintain the likelihood ratio of generated sequences. Using DENs, we engineered polyadenylation signals with more than 10-fold higher selection odds than the best gradient ascent-generated patterns, identified splice regulatory sequences predicted to result in highly differential splicing between cell lines, and improved on state-of-the-art results for protein design tasks.

MISCAST: MIssense variant to protein StruCture Analysis web SuiTe.

Human genome sequencing efforts have greatly expanded, and a plethora of missense variants identified both in patients and in the general population is now publicly accessible. Interpretation of the molecular-level effect of missense variants, however, remains challenging and requires a particular investigation of amino acid substitutions in the context of protein structure and function. Answers to questions like ‘Is a variant perturbing a site involved in key macromolecular interactions and/or cellular signaling?’, or ‘Is a variant changing an amino acid located at the protein core or part of a cluster of known pathogenic mutations in 3D?’ are crucial. Motivated by these needs, we developed MISCAST (missense variant to protein structure analysis web suite; http://miscast.broadinstitute.org/). MISCAST is an interactive and user-friendly web server to visualize and analyze missense variants in protein sequence and structure space. Additionally, a comprehensive set of protein structural and functional features have been aggregated in MISCAST from multiple databases, and displayed on structures alongside the variants to provide users with the biological context of the variant location in an integrated platform. We further made the annotated data and protein structures readily downloadable from MISCAST to foster advanced offline analysis of missense variants by a wide biological community.

Constraints on the expansion of paralogous protein families.

Duplication and divergence is a major mechanism by which new proteins and functions emerge in biology. Consequently, most organisms, in all domains of life, have genomes that encode large paralogous families of proteins. For recently duplicated pathways to acquire different, independent functions, the two paralogs must acquire mutations that effectively insulate them from one another. For instance, paralogous signaling proteins must acquire mutations that endow them with different interaction specificities such that they can participate in different signaling pathways without disruptive cross-talk. Although duplicated genes undoubtedly shape each other's evolution as they diverge and attain new functions, it is less clear how other paralogs impact or constrain gene duplication. Does the establishment of a new pathway by duplication and divergence require the system-wide optimization of all paralogs? The answer has profound implications for molecular evolution and our ability to engineer biological systems. Here, we discuss models, experiments, and approaches for tackling this question, and for understanding how new proteins and pathways are born.

Variant effect predictions capture some aspects of deep mutational scanning experiments

BackgroundDeep mutational scanning (DMS) studies exploit the mutational landscape of sequence variation by systematically and comprehensively assaying the effect of single amino acid variants (SAVs; also referred to as missense mutations, or non-synonymous Single Nucleotide Variants – missense SNVs or nsSNVs) for particular proteins. We assembled SAV annotations from 22 different DMS experiments and normalized the effect scores to evaluate variant effect prediction methods. Three trained on traditional variant effect data (PolyPhen-2, SIFT, SNAP2), a regression method optimized on DMS data (Envision), and a naïve prediction using conservation information from homologs.ResultsOn a set of 32,981 SAVs, all methods captured some aspects of the experimental effect scores, albeit not the same. Traditional methods such as SNAP2 correlated slightly more with measurements and better classified binary states (effect or neutral). Envision appeared to better estimate the precise degree of effect. Most surprising was that the simple naïve conservation approach using PSI-BLAST in many cases outperformed other methods. All methods captured beneficial effects (gain-of-function) significantly worse than deleterious (loss-of-function). For the few proteins with multiple independent experimental measurements, experiments differed substantially, but agreed more with each other than with predictions.ConclusionsDMS provides a new powerful experimental means of understanding the dynamics of the protein sequence space. As always, promising new beginnings have to overcome challenges. While our results demonstrated that DMS will be crucial to improve variant effect prediction methods, data diversity hindered simplification and generalization.

Structure and function of Bs164 β-mannosidase from Bacteroides salyersiae the founding member of glycoside hydrolase family GH164

Recent work exploring protein sequence space has revealed a new glycoside hydrolase (GH) family (GH164) of putative mannosidases. GH164 genes are present in several commensal bacteria, implicating these genes in the degradation of dietary glycans. However, little is known about the structure, mechanism of action, and substrate specificity of these enzymes. Herein we report the biochemical characterization and crystal structures of the founding member of this family (Bs164) from the human gut symbiont Bacteroides salyersiae. Previous reports of this enzyme indicated that it has α-mannosidase activity, however, we conclusively show that it cleaves only β-mannose linkages. Using NMR spectroscopy, detailed enzyme kinetics of WT and mutant Bs164, and multiangle light scattering we found that it is a trimeric retaining β-mannosidase, that is susceptible to several known mannosidase inhibitors. X-ray crystallography revealed the structure of Bs164, the first known structure of a GH164, at 1.91 Å resolution. Bs164 is composed of three domains: a (β/α)8 barrel, a trimerization domain, and a β-sandwich domain, representing a previously unobserved structural-fold for β-mannosidases. Structures of Bs164 at 1.80–2.55 Å resolution in complex with the inhibitors noeuromycin, mannoimidazole, or 2,4-dinitrophenol 2-deoxy-2-fluoro-mannoside reveal the residues essential for specificity and catalysis including the catalytic nucleophile (Glu-297) and acid/base residue (Glu-160). These findings further our knowledge of the mechanisms commensal microbes use for nutrient acquisition.

Multi-state design of flexible proteins predicts sequences optimal for conformational change.

Computational protein design of an ensemble of conformations for one protein-i.e., multi-state design-determines the side chain identity by optimizing the energetic contributions of that side chain in each of the backbone conformations. Sampling the resulting large sequence-structure search space limits the number of conformations and the size of proteins in multi-state design algorithms. Here, we demonstrated that the REstrained CONvergence (RECON) algorithm can simultaneously evaluate the sequence of large proteins that undergo substantial conformational changes. Simultaneous optimization of side chain conformations across all conformations increased sequence conservation when compared to single-state designs in all cases. More importantly, the sequence space sampled by RECON MSD resembled the evolutionary sequence space of flexible proteins, particularly when confined to predicting the mutational preferences of limited common ancestral descent, such as in the case of influenza type A hemagglutinin. Additionally, we found that sequence positions which require substantial changes in their local environment across an ensemble of conformations are more likely to be conserved. These increased conservation rates are better captured by RECON MSD over multiple conformations and thus multiple local residue environments during design. To quantify this rewiring of contacts at a certain position in sequence and structure, we introduced a new metric designated 'contact proximity deviation' that enumerates contact map changes. This measure allows mapping of global conformational changes into local side chain proximity adjustments, a property not captured by traditional global similarity metrics such as RMSD or local similarity metrics such as changes in φ and ψ angles.

Identification and Prediction of Fold-Switching Proteins

A central tenet of biology is that globular proteins have a unique three-dimensional structure under physiological conditions. Over the past 15 years, however, a growing number of proteins have been shown to switch folds, i.e. remodel their secondary and tertiary structures in response to cellular stimuli, leading to biologically relevant functional changes. Previously, we identified 96 of these fold-switching proteins in the Protein Data Bank (PDB) [1]. We also found evidence suggesting that their natural abundance likely exceeds their PDB representation. To gauge how pervasive fold switchers might be in nature, we sought to explore their evolutionary conservation and sequence properties by developing a bioinformatics-based method to predict previously unidentified fold switchers from their genomic sequences. We used this method to perform blind predictions on regions of protein sequence space known to contain both fold switchers and non-fold switchers. This space comprises ∼7000 non-redundant sequences from a family of bacterial transcription factors known as RfaH and NusG. RfaH regulates protein expression by functioning as both a transcription factor with an alpha-helical fold and a translation factor with a beta-barrel fold. In contrast, NusG maintains its beta-barrel fold and functions as a transcription factor. Upon mapping RfaH/NusG sequence space, we predicted the propensities of several sequences therein to switch folds. Our predictions, validated by several experimental methods, suggest that fold switching is conserved among distantly-related bacterial phyla, highlighting its evolutionary robustness. Furthermore, preliminary evidence indicates that sequence constraints are looser for fold-switching sequences than for non-fold-switching ones, suggesting that many uncharacterized genomic sequences could potentially switch folds. [1] Porter and Looger (2018) Proc Natl Acad Sci USA 115:5968-5973

A Strategy for Combinatorial Cavity Design in De Novo Proteins.

Protein sequence space is vast; nature uses only an infinitesimal fraction of possible sequences to sustain life. Are there solutions to biological problems other than those provided by nature? Can we create artificial proteins that sustain life? To investigate these questions, we have created combinatorial collections, or libraries, of novel sequences with no homology to those found in living organisms. Previously designed libraries contained numerous functional proteins. However, they often formed dynamic, rather than well-ordered structures, which complicated structural and mechanistic characterization. To address this challenge, we describe the development of new libraries based on the de novo protein S-824, a 4-helix bundle with a very stable 3-dimensional structure. Distinct from previous libraries, we targeted variability to a specific region of the protein, seeking to create potential functional sites. By characterizing variant proteins from this library, we demonstrate that the S-824 scaffold tolerates diverse amino acid substitutions in a putative cavity, including buried polar residues suitable for catalysis. We designed and created a DNA library encoding 1.7 × 106 unique protein sequences. This new library of stable de novo α-helical proteins is well suited for screens and selections for a range of functional activities in vitro and in vivo.

Machine learning-assisted enzyme engineering.

Directed evolution and rational design are powerful strategies in protein engineering to tailor enzyme properties to meet the demands in academia and industry. Traditional approaches for enzyme engineering and directed evolution are often experimentally driven, in particular when the protein structure-function relationship is not available. Though they have been successfully applied to engineer many enzymes, these methods are still facing significant challenges due to the tremendous size of the protein sequence space and the combinatorial problem. It can be ascertained that current experimental techniques and computational techniques might never be able to sample through the entire protein sequence space and benefit from nature's full potential for the generation of better enzymes. With advancements in next generation sequencing, high throughput screening methods, the growth of protein databases and artificial intelligence, especially machine learning (ML), data-driven enzyme engineering is emerging as a promising solution to these challenges. To date, ML-assisted approaches have efficiently and accurately determined the quantitative structure-property/activity relationship for the prediction of diverse enzyme properties. In addition, enzyme engineering can be accelerated much faster than ever through the combination of experimental library generation and ML-based prediction. In this chapter, we review the recent progresses in ML-assisted enzyme engineering and highlight several successful examples (e.g., to enhance activity, enantioselectivity, or thermostability). Herein we explain enzyme engineering strategies that combine random or (semi-)rational approaches with ML methods and allow an effective reengineering of enzymes to improve targeted properties. We further discuss the main challenges to solve in order to realize the full potential of ML methods in enzyme engineering. Finally, we describe the current limitations of ML-assisted enzyme engineering, and our perspective on future opportunities in this growing field.

What limits the primary sequence space of natural proteins?

Number of naturally occurring primary sequences of proteins is an infinitesimally small subset of the possible number of primary sequences that can be synthesized using 20 amino acids. Prevailing views ascribe this to slow and incremental mutational/selection evolutionary mechanisms. However, considering the large number of avenues available in form of diversity of emerging/evolving and/or disappearing living systems for exploring the primary sequence space over the evolutionary time scale of ∼3.5 billion years, this remains a conjecture. Therefore, to investigate primary sequence space limitations, we carried out a systematic study for finding primary sequences absent in nature. We report the discovery of the smallest peptide sequence “Cysteine-Glutamine-Tryptophan-Tryptophan” that is not found in over half-a-million curated protein sequences in the Uniprot (Swiss-Prot) database. Additionally, we report a library of 83605 pentapeptides that are not found in any of the known protein sequences. Compositional analyses of these absent primary sequences yield a remarkably strong power relationship between the percentage occurrence of individual amino acids in all known protein sequences and their respective frequency of occurrence in the absent peptides, regardless of their specific position in the sequences. If random evolutionary mechanisms were responsible for limitations to the primary sequence space, then one would not expect any relationship between compositions of available and absent primary sequences. Thus, we conclusively show that stoichiometric constraints on amino acids limit the primary sequence space of proteins in nature. We discuss the possibly profound implications of our findings in both evolutionary and synthetic biology.Communicated by Ramaswamy H. Sarma

Size and structure of the sequence space of repeat proteins.

The coding space of protein sequences is shaped by evolutionary constraints set by requirements of function and stability. We show that the coding space of a given protein family—the total number of sequences in that family—can be estimated using models of maximum entropy trained on multiple sequence alignments of naturally occuring amino acid sequences. We analyzed and calculated the size of three abundant repeat proteins families, whose members are large proteins made of many repetitions of conserved portions of ∼30 amino acids. While amino acid conservation at each position of the alignment explains most of the reduction of diversity relative to completely random sequences, we found that correlations between amino acid usage at different positions significantly impact that diversity. We quantified the impact of different types of correlations, functional and evolutionary, on sequence diversity. Analysis of the detailed structure of the coding space of the families revealed a rugged landscape, with many local energy minima of varying sizes with a hierarchical structure, reminiscent of fustrated energy landscapes of spin glass in physics. This clustered structure indicates a multiplicity of subtypes within each family, and suggests new strategies for protein design.

Visualization of protein sequence space with force-directed graphs, and their application to the choice of target-template pairs for homology modelling

The protein sequence-structure gap results from the contrast between rapid, low-cost deep sequencing, and slow, expensive experimental structure determination techniques. Comparative homology modelling may have the potential to close this gap by predicting protein structure in target sequences using existing experimentally solved structures as templates. This paper presents the first use of force-directed graphs for the visualization of sequence space in two dimensions, and applies them to the choice of suitable RNA-dependent RNA polymerase (RdRP) target-template pairs within human-infective RNA virus genera. Measures of centrality in protein sequence space for each genus were also derived and used to identify centroid nearest-neighbour sequences (CNNs) potentially useful for production of homology models most representative of their genera. Homology modelling was then carried out for target-template pairs in different species, different genera and different families, and model quality assessed using several metrics. Reconstructed ancestral RdRP sequences for individual genera were also used as templates for the production of ancestral RdRP homology models. High quality ancestral RdRP models were consistently produced, as were good quality models for target-template pairs in the same genus. Homology modelling between genera in the same family produced mixed results and inter-family modelling was unreliable. We present a protocol for the production of optimal RdRP homology models for use in further experiments, e.g. docking to discover novel anti-viral compounds. (219 words)

Evolutionary couplings of amino acid residues reveal structure and function of bacterial signaling proteins.

The genomic era along with major advances in high-throughput sequencing technology has led to a rapid expansion of the genomic and consequently the protein sequence space. Bacterial extracytoplasmic function sigma factors have emerged as an important group of signaling proteins in bacteria involved in many regulatory decisions, most notably the adaptation to cell envelope stress. Their wide prevalence and amplification among bacterial genomes has led to sub-group classification and the realization of diverse signaling mechanisms. Mathematical frameworks have been developed to utilize extensive protein sequence alignments to extract co-evolutionary signals of interaction. This has proven useful in a number of different biological fields, including de novo structure prediction, protein-protein partner identification and the elucidation of alternative protein conformations for signal proteins, to name a few. The mathematical tools, commonly referred to under the name 'Direct Coupling Analysis' have now been applied to deduce molecular mechanisms of activation for sub-groups of extracytoplasmic sigma factors adding to previous successes on bacterial two-component signaling proteins. The amplification of signal transduction protein genes in bacterial genomes made them the first to be amenable to this approach but the sequences are available now to aid the molecular microbiologist, no matter their protein pathway of interest.

ProteinNet: a standardized data set for machine learning of protein structure

BackgroundRapid progress in deep learning has spurred its application to bioinformatics problems including protein structure prediction and design. In classic machine learning problems like computer vision, progress has been driven by standardized data sets that facilitate fair assessment of new methods and lower the barrier to entry for non-domain experts. While data sets of protein sequence and structure exist, they lack certain components critical for machine learning, including high-quality multiple sequence alignments and insulated training/validation splits that account for deep but only weakly detectable homology across protein space.ResultsWe created the ProteinNet series of data sets to provide a standardized mechanism for training and assessing data-driven models of protein sequence-structure relationships. ProteinNet integrates sequence, structure, and evolutionary information in programmatically accessible file formats tailored for machine learning frameworks. Multiple sequence alignments of all structurally characterized proteins were created using substantial high-performance computing resources. Standardized data splits were also generated to emulate the difficulty of past CASP (Critical Assessment of protein Structure Prediction) experiments by resetting protein sequence and structure space to the historical states that preceded six prior CASPs. Utilizing sensitive evolution-based distance metrics to segregate distantly related proteins, we have additionally created validation sets distinct from the official CASP sets that faithfully mimic their difficulty.ConclusionProteinNet represents a comprehensive and accessible resource for training and assessing machine-learned models of protein structure.

GeneORator: An Effective Strategy for Navigating Protein Sequence Space More Efficiently through Boolean OR-Type DNA Libraries.

Directed evolution requires the creation of genetic diversity and subsequent screening or selection for improved variants. For DNA mutagenesis, conventional site-directed methods implicitly utilize the Boolean AND operator (creating all mutations simultaneously), producing a combinatorial explosion in the number of genetic variants as the number of mutations increases. We introduce GeneORator, a novel strategy for creating DNA libraries based on the Boolean logical OR operator. Here, a single library is divided into many subsets, each containing different combinations of the desired mutations. Consequently, the effect of adding more mutations on the number of genetic combinations is additive (Boolean OR logic) and not exponential (AND logic). We demonstrate this strategy with large-scale mutagenesis studies, using monoamine oxidase-N (Aspergillus niger) as the exemplar target. First, we mutated every residue in the secondary structure-containing regions (276 out of a total 495 amino acids) to screen for improvements in kcat. Second, combinatorial OR-type libraries permitted screening of diverse mutation combinations in the enzyme active site to detect activity toward novel substrates. In both examples, OR-type libraries effectively reduced the number of variants searched up to 1010-fold, dramatically reducing the screening effort required to discover variants with improved and/or novel activity. Importantly, this approach enables the screening of a greater diversity of mutation combinations, accessing a larger area of a protein’s sequence space. OR-type libraries can be applied to any biological engineering objective requiring DNA mutagenesis, and the approach has wide ranging applications in, for example, enzyme engineering, antibody engineering, and synthetic biology.

Understanding molecular mechanisms in cell signaling through natural and artificial sequence variation

The functionally tolerated sequence space of proteins can now be explored in an unprecedented way, owing to the expansion of genomic databases and the development of high-throughput methods to interrogate protein function. For signaling proteins, several recent studies have shown how the analysis of sequence variation leverages the available protein-structure information to provide new insights into specificity and allosteric regulation. In this Review, we discuss recent work that illustrates how this emerging approach is providing a deeper understanding of signaling proteins.

Design of metalloproteins and novel protein folds using variational autoencoders

The design of novel proteins has many applications but remains an attritional process with success in isolated cases. Meanwhile, deep learning technologies have exploded in popularity in recent years and are increasingly applicable to biology due to the rise in available data. We attempt to link protein design and deep learning by using variational autoencoders to generate protein sequences conditioned on desired properties. Potential copper and calcium binding sites are added to non-metal binding proteins without human intervention and compared to a hidden Markov model. In another use case, a grammar of protein structures is developed and used to produce sequences for a novel protein topology. One candidate structure is found to be stable by molecular dynamics simulation. The ability of our model to confine the vast search space of protein sequences and to scale easily has the potential to assist in a variety of protein design tasks.

Selection for Protein Stability Enriches for Epistatic Interactions.

A now classical argument for the marginal thermodynamic stability of proteins explains the distribution of observed protein stabilities as a consequence of an entropic pull in protein sequence space. In particular, most sequences that are sufficiently stable to fold will have stabilities near the folding threshold. Here, we extend this argument to consider its predictions for epistatic interactions for the effects of mutations on the free energy of folding. Although there is abundant evidence to indicate that the effects of mutations on the free energy of folding are nearly additive and conserved over evolutionary time, we show that these observations are compatible with the hypothesis that a non-additive contribution to the folding free energy is essential for observed proteins to maintain their native structure. In particular, through both simulations and analytical results, we show that even very small departures from additivity are sufficient to drive this effect.

A quantitative map of protein sequence space for the cis-defensin superfamily.

The cis-defensins are a superfamily of small, cationic, cysteine-rich proteins, sharing a common scaffold, but highly divergent sequences and varied functions from host-defence to signalling. Superfamily members are most abundant in plants (with some genomes containing hundreds of members), but are also found across fungi and invertebrates. However, of the thousands of cis-defensin sequences in databases, only have a handful have solved structures or assigned activities. Non-phylogenetic sequence-analysis methods are therefore necessary to use the relationships within the superfamily to classify members, and to predict and engineer functions. We show that the generation of a quantitative map of sequence space allows these highly divergent sequences to be usefully analyzed. This information-rich technique can identify natural groupings of sequences with similar biophysical properties, detect interpretable covarying properties, and provide information on typical or intermediate sequences for each cluster. The cis-defensin superfamily contains clearly-defined groups, identifiable based on their biophysical properties and motifs. The organization of sequences within this space also provides a foundation of understanding the ancient evolution of the superfamily. A webtool for exploring and querying the space is hosted at TS404.shinyapps.io/DefSpace. Supplementary data are available at Bioinformatics online.