Illuminating protein space with a programmable generative model.

Three billion years of evolution has produced a tremendous diversity of protein molecules1, but the full potential of proteins is likely to be much greater. Accessing this potential has been challenging for both computation and experiments because the space of possible protein molecules is much larger than the space of those likely to have functions. Here we introduce Chroma, a generative model for proteins and protein complexes that can directly sample novel protein structures and sequences, and that can be conditioned to steer the generative process towards desired properties and functions. To enable this, we introduce a diffusion process that respects the conformational statistics of polymer ensembles, an efficient neural architecture for molecular systems that enables long-range reasoning with sub-quadratic scaling, layers for efficiently synthesizing three-dimensional structures of proteins from predicted inter-residue geometries and a general low-temperature sampling algorithm for diffusion models. Chroma achieves protein design as Bayesian inference under external constraints, which can involve symmetries, substructure, shape, semantics and even natural-language prompts. The experimental characterization of 310 proteins shows that sampling from Chroma results in proteins that are highly expressed, fold and have favourable biophysical properties. The crystal structures of two designed proteins exhibit atomistic agreement with Chroma samples (a backbone root-mean-square deviation of around 1.0 Å). With this unified approach to protein design, we hope to accelerate the programming of protein matter to benefit human health, materials science and synthetic biology.

Sequence-sensitive elastic network captures dynamical features necessary for miR-125a maturation.

The Elastic Network Contact Model (ENCoM) is a coarse-grained normal mode analysis (NMA) model unique in its all-atom sensitivity to the sequence of the studied macromolecule and thus to the effect of mutations. We adapted ENCoM to simulate the dynamics of ribonucleic acid (RNA) molecules, benchmarked its performance against other popular NMA models and used it to study the 3D structural dynamics of human microRNA miR-125a, leveraging high-throughput experimental maturation efficiency data of over 26 000 sequence variants. We also introduce a novel way of using dynamical information from NMA to train multivariate linear regression models, with the purpose of highlighting the most salient contributions of dynamics to function. ENCoM has a similar performance profile on RNA than on proteins when compared to the Anisotropic Network Model (ANM), the most widely used coarse-grained NMA model; it has the advantage on predicting large-scale motions while ANM performs better on B-factors prediction. A stringent benchmark from the miR-125a maturation dataset, in which the training set contains no sequence information in common with the testing set, reveals that ENCoM is the only tested model able to capture signal beyond the sequence. This ability translates to better predictive power on a second benchmark in which sequence features are shared between the train and test sets. When training the linear regression model using all available data, the dynamical features identified as necessary for miR-125a maturation point to known patterns but also offer new insights into the biogenesis of microRNAs. Our novel approach combining NMA with multivariate linear regression is generalizable to any macromolecule for which relatively high-throughput mutational data is available.

Data-driven computational protein design.

Computational protein design can generate proteins not found in nature that adopt desired structures and perform novel functions. Although proteins could, in theory, be designed with ab initio methods, practical success has come from using large amounts of data that describe the sequences, structures, and functions of existing proteins and their variants. We present recent creative uses of multiple-sequence alignments, protein structures, and high-throughput functional assays in computational protein design. Approaches range from enhancing structure-based design with experimental data to building regression models to training deep neural nets that generate novel sequences. Looking ahead, deep learning will be increasingly important for maximizing the value of data for protein design.

Tertiary Structural Motif Sequence Statistics Enable Facile Prediction and Design of Peptides that Bind Anti-apoptotic Bfl-1 and Mcl-1.

Understanding the relationship between protein sequence and structure well enough to design new proteins with desired functions is a longstanding goal in protein science. Here, we show that recurring tertiary structural motifs (TERMs) in the PDB provide rich information for protein-peptide interaction prediction and design. TERM statistics can be used to predict peptide binding energies for Bcl-2 family proteins as accurately as widely used structure-based tools. Furthermore, design using TERM energies (dTERMen) rapidly and reliably generates high-affinity peptide binders of anti-apoptotic proteins Bfl-1 and Mcl-1 with just 15%-38% sequence identity to any known native Bcl-2 family protein ligand. High-resolution structures of four designed peptides bound to their targets provide opportunities to analyze the strengths and limitations of the computational design method. Our results support dTERMen as a powerful approach that can complement existing tools for protein engineering.

PixelDB: Protein-peptide complexes annotated with structural conservation of the peptide binding mode.

PixelDB, the Peptide Exosite Location Database, compiles 1966 non-redundant, high-resolution structures of protein-peptide complexes filtered to minimize the impact of crystal packing on peptide conformation. The database is organized to facilitate study of structurally conserved versus non-conserved elements of protein-peptide engagement. PixelDB clusters complexes based on the structural similarity of the peptide-binding protein, and by comparing complexes within a cluster highlights examples of domains that engage peptides using more than one binding mode. PixelDB also identifies conserved peptide core structural motifs characteristic of each binding mode. Peptide regions that flank core motifs often make non-structurally conserved interactions with the protein surface in regions we call exosites. Many examples establish that exosite contacts can be important for enhancing protein binding and interaction specificity. PixelDB provides a resource for computational and structural biologists to study, model, and predict core-motif and exosite-contacting peptide interactions. PixelDB is available to the community without restriction in a convenient flat-file format with accompanying visualization tools.

Mutations that Allow SIR2 Orthologs to Function in a NAD+-Depleted Environment.

Sirtuin enzymes depend on NAD+ to catalyze protein deacetylation. Therefore, the lowering of NAD+ during aging leads to decreased sirtuin activity and may speed up aging processes in laboratory animals and humans. In this study, we used a genetic screen to identify two mutations in the catalytic domain of yeast Sir2 that allow the enzyme to function in an NAD+-depleted environment. These mutant enzymes give rise to a significant increase of yeast replicative lifespan and increase deacetylation by the Sir2 ortholog, SIRT1, in mammalian cells. Our data suggest that these mutations increase the stability of the conserved catalytic sirtuin domain, thereby increasing the catalytic efficiency of the mutant enzymes. Our approach to identifying sirtuin mutants that permit function in NAD+-limited environments may inform the design of small molecules that can maintain sirtuin activity in aging organisms.

Applications of Normal Mode Analysis Methods in Computational Protein Design.

Recent advances in coarse-grained normal mode analysis methods make possible the large-scale prediction of the effect of mutations on protein stability and dynamics as well as the generation of biologically relevant conformational ensembles. Given the interplay between flexibility and enzymatic activity, the combined analysis of stability and dynamics using the Elastic Network Contact Model (ENCoM) method has ample applications in protein engineering in industrial and medical applications such as in computational antibody design. Here, we present a detailed tutorial on how to perform such calculations using ENCoM.

ENCoM server: exploring protein conformational space and the effect of mutations on protein function and stability.

ENCoM is a coarse-grained normal mode analysis method recently introduced that unlike previous such methods is unique in that it accounts for the nature of amino acids. The inclusion of this layer of information was shown to improve conformational space sampling and apply for the first time a coarse-grained normal mode analysis method to predict the effect of single point mutations on protein dynamics and thermostability resulting from vibrational entropy changes. Here we present a web server that allows non-technical users to have access to ENCoM calculations to predict the effect of mutations on thermostability and dynamics as well as to generate geometrically realistic conformational ensembles. The server is accessible at: http://bcb.med.usherbrooke.ca/encom.

Vibrational entropy differences between mesophile and thermophile proteins and their use in protein engineering.

We recently introduced ENCoM, an elastic network atomic contact model, as the first coarse-grained normal mode analysis method that accounts for the nature of amino acids and can predict the effect of mutations on thermostability based on changes vibrational entropy. In this proof-of-concept article, we use pairs of mesophile and thermophile homolog proteins with identical structures to determine if a measure of vibrational entropy based on normal mode analysis can discriminate thermophile from mesophile proteins. We observe that in around 60% of cases, thermophile proteins are more rigid at equivalent temperatures than their mesophile counterpart and this difference can guide the design of proteins to increase their thermostability through series of mutations. We observe that mutations separating thermophile proteins from their mesophile orthologs contribute independently to a decrease in vibrational entropy and discuss the application and implications of this methodology to protein engineering.

A coarse-grained elastic network atom contact model and its use in the simulation of protein dynamics and the prediction of the effect of mutations.

Normal mode analysis (NMA) methods are widely used to study dynamic aspects of protein structures. Two critical components of NMA methods are coarse-graining in the level of simplification used to represent protein structures and the choice of potential energy functional form. There is a trade-off between speed and accuracy in different choices. In one extreme one finds accurate but slow molecular-dynamics based methods with all-atom representations and detailed atom potentials. On the other extreme, fast elastic network model (ENM) methods with Cα-only representations and simplified potentials that based on geometry alone, thus oblivious to protein sequence. Here we present ENCoM, an Elastic Network Contact Model that employs a potential energy function that includes a pairwise atom-type non-bonded interaction term and thus makes it possible to consider the effect of the specific nature of amino-acids on dynamics within the context of NMA. ENCoM is as fast as existing ENM methods and outperforms such methods in the generation of conformational ensembles. Here we introduce a new application for NMA methods with the use of ENCoM in the prediction of the effect of mutations on protein stability. While existing methods are based on machine learning or enthalpic considerations, the use of ENCoM, based on vibrational normal modes, is based on entropic considerations. This represents a novel area of application for NMA methods and a novel approach for the prediction of the effect of mutations. We compare ENCoM to a large number of methods in terms of accuracy and self-consistency. We show that the accuracy of ENCoM is comparable to that of the best existing methods. We show that existing methods are biased towards the prediction of destabilizing mutations and that ENCoM is less biased at predicting stabilizing mutations.

The mode of action of recombinant Mycobacterium tuberculosis shikimate kinase: kinetics and thermodynamics analyses.

Tuberculosis remains as one of the main cause of mortality worldwide due to a single infectious agent, Mycobacterium tuberculosis. The aroK-encoded M. tuberculosis Shikimate Kinase (MtSK), shown to be essential for survival of bacilli, catalyzes the phosphoryl transfer from ATP to the carbon-3 hydroxyl group of shikimate (SKH), yielding shikimate-3-phosphate and ADP. Here we present purification to homogeneity, and oligomeric state determination of recombinant MtSK. Biochemical and biophysical data suggest that the chemical reaction catalyzed by monomeric MtSK follows a rapid-equilibrium random order of substrate binding, and ordered product release. Isothermal titration calorimetry (ITC) for binding of ligands to MtSK provided thermodynamic signatures of non-covalent interactions to each process. A comparison of steady-state kinetics parameters and equilibrium dissociation constant value determined by ITC showed that ATP binding does not increase the affinity of MtSK for SKH. We suggest that MtSK would more appropriately be described as an aroL-encoded type II shikimate kinase. Our manuscript also gives thermodynamic description of SKH binding to MtSK and data for the number of protons exchanged during this bimolecular interaction. The negative value for the change in constant pressure heat capacity (ΔCp) and molecular homology model building suggest a pronounced contribution of desolvation of non-polar groups upon binary complex formation. Thermodynamic parameters were deconvoluted into hydrophobic and vibrational contributions upon MtSK:SKH binary complex formation. Data for the number of protons exchanged during this bimolecular interaction are interpreted in light of a structural model to try to propose the likely amino acid side chains that are the proton donors to bulk solvent following MtSK:SKH complex formation.

StAR-related lipid transfer domain protein 5 binds primary bile acids.

Steroidogenic acute regulatory-related lipid transfer (START) domain proteins are involved in the nonvesicular intracellular transport of lipids and sterols. The STARD1 (STARD1 and STARD3) and STARD4 subfamilies (STARD4-6) have an internal cavity large enough to accommodate sterols. To provide a deeper understanding on the structural biology of this domain, the binding of sterols to STARD5, a member of the STARD4 subfamily, was monitored. The SAR by NMR [(1)H-(15)N heteronuclear single-quantum coherence (HSQC)] approach, complemented by circular dichroism (CD) and isothermal titration calorimetry (ITC), was used. Titration of STARD5 with cholic (CA) and chenodeoxycholic acid (CDCA), ligands of the farnesoid X receptor (FXR), leads to drastic perturbation of the (1)H-(15)N HSQC spectra and the identification of the residues in contact with those ligands. The most perturbed residues in presence of ligands are lining the internal cavity of the protein. Ka values of 1.8·10-(4) M(-1) and 6.3·10(4) M(-1) were measured for CA and CDCA, respectively. This is the first report of a START domain protein in complex with a sterol ligand. Our original findings indicate that STARD5 may be involved in the transport of bile acids rather than cholesterol.

New structural determinants for c-Myc specific heterodimerization with Max and development of a novel homodimeric c-Myc b-HLH-LZ.

c-Myc must heterodimerize with Max to accomplish its functions as a transcription factor. This specific heterodimerization occurs through the b-HLH-LZ (basic region, helix 1-loop-helix 2-leucine zipper) domains. In fact, many studies have shown that the c-Myc b-HLH-LZ (c-Myc'SH) preferentially forms a heterodimer with the Max b-HLH-LZ (Max'SH). The primary mechanism underlying the specific heterodimerization lies on the destabilization of both homodimers and the formation of a more stable heterodimer. In this regard, it has been widely reported that c-Myc'SH has low solubility and homodimerizes poorly and that repulsions within the LZ domain account for the homodimer instability. Here, we show that replacing one residue in the basic region and one residue in Helix 1 (H(1)) of c-Myc'SH with corresponding residues conserved in b-HLH proteins confers to c-Myc'SH a higher propensity to form a stable homodimer in solution. In stark contrast to the wild-type protein, this double mutant (L362R, R367L) of the c-Myc b-HLH-LZ (c-Myc'RL) shows limited heterodimerization with Max'SH in vitro. In addition, c-Myc'RL forms highly stable and soluble complexes with canonical as well as non-canonical E-box probes. Altogether, our results demonstrate for the first time that structural determinants driving the specific heterodimerization of c-Myc and Max are embedded in the basic region and H(1) of c-Myc and that these can be exploited to engineer a novel homodimeric c-Myc b-HLH-LZ with the ability of binding the E-box sequence autonomously and with high affinity.