A computational framework to empower probabilistic protein design

De Novo Protein Design for Novel Folds Using Guided Conditional Wasserstein Generative Adversarial Networks

Although massive data is quickly accumulating on protein sequence and structure, there is a small and limited number of protein architectural types (or structural folds). This study is addressing the following question: how well could one reveal underlying sequence-structure relationships and design protein sequences for an arbitrary, potentially novel, structural fold? In response to the question, we have developed novel deep generative models, namely, semisupervised gcWGAN (guided, conditional, Wasserstein Generative Adversarial Networks). To overcome training difficulties and improve design qualities, we build our models on conditional Wasserstein GAN (WGAN) that uses Wasserstein distance in the loss function. Our major contributions include (1) constructing a low-dimensional and generalizable representation of the fold space for the conditional input, (2) developing an ultrafast sequence-to-fold predictor (or oracle) and incorporating its feedback into WGAN as a loss to guide model training, and (3) exploiting sequence data with and without paired structures to enable a semisupervised training strategy. Assessed by the oracle over 100 novel folds not in the training set, gcWGAN generates more successful designs and covers 3.5 times more target folds compared to a competing data-driven method (cVAE). Assessed by sequence- and structure-based predictors, gcWGAN designs are physically and biologically sound. Assessed by a structure predictor over representative novel folds, including one not even part of basis folds, gcWGAN designs have comparable or better fold accuracy yet much more sequence diversity and novelty than cVAE. The ultrafast data-driven model is further shown to boost the success of a principle-driven de novo method (RosettaDesign), through generating design seeds and tailoring design space. In conclusion, gcWGAN explores uncharted sequence space to design proteins by learning generalizable principles from current sequence-structure data. Data, source codes, and trained models are available at https://github.com/Shen-Lab/gcWGAN.

iCFN: an efficient exact algorithm for multistate protein design

Motivation

Multistate protein design addresses real-world challenges, such as multi-specificity design and backbone flexibility, by considering both positive and negative protein states with an ensemble of substates for each. It also presents an enormous challenge to exact algorithms that guarantee the optimal solutions and enable a direct test of mechanistic hypotheses behind models. However, efficient exact algorithms are lacking for multistate protein design.

Results

We have developed an efficient exact algorithm called interconnected cost function networks (iCFN) for multistate protein design. Its generic formulation allows for a wide array of applications such as stability, affinity and specificity designs while addressing concerns such as global flexibility of protein backbones. iCFN treats each substate design as a weighted constraint satisfaction problem (WCSP) modeled through a CFN; and it solves the coupled WCSPs using novel bounds and a depth-first branch-and-bound search over a tree structure of sequences, substates, and conformations. When iCFN is applied to specificity design of a T-cell receptor, a problem of unprecedented size to exact methods, it drastically reduces search space and running time to make the problem tractable. Moreover, iCFN generates experimentally-agreeing receptor designs with improved accuracy compared with state-of-the-art methods, highlights the importance of modeling backbone flexibility in protein design, and reveals molecular mechanisms underlying binding specificity.

Availability and implementation

https://shen-lab.github.io/software/iCFN.

Supplementary information

Supplementary data are available at Bioinformatics online.

BBK* (Branch and Bound Over K*): A Provable and Efficient Ensemble-Based Protein Design Algorithm to Optimize Stability and Binding Affinity Over Large Sequence Spaces

Computational protein design (CPD) algorithms that compute binding affinity, K_a, search for sequences with an energetically favorable free energy of binding. Recent work shows that three principles improve the biological accuracy of CPD: ensemble-based design, continuous flexibility of backbone and side-chain conformations, and provable guarantees of accuracy with respect to the input. However, previous methods that use all three design principles are single-sequence (SS) algorithms, which are very costly: linear in the number of sequences and thus exponential in the number of simultaneously mutable residues. To address this computational challenge, we introduce BBK*, a new CPD algorithm whose key innovation is the multisequence (MS) bound: BBK* efficiently computes a single provable upper bound to approximate K_a for a combinatorial number of sequences, and avoids SS computation for all provably suboptimal sequences. Thus, to our knowledge, BBK* is the first provable, ensemble-based CPD algorithm to run in time sublinear in the number of sequences. Computational experiments on 204 protein design problems show that BBK* finds the tightest binding sequences while approximating K_a for up to 10⁵-fold fewer sequences than the previous state-of-the-art algorithms, which require exhaustive enumeration of sequences. Furthermore, for 51 protein-ligand design problems, BBK* provably approximates K_a up to 1982-fold faster than the previous state-of-the-art iMinDEE/[Formula: see text]/[Formula: see text] algorithm. Therefore, BBK* not only accelerates protein designs that are possible with previous provable algorithms, but also efficiently performs designs that are too large for previous methods.

A critical analysis of computational protein design with sparse residue interaction graphs

Protein design algorithms enumerate a combinatorial number of candidate structures to compute the Global Minimum Energy Conformation (GMEC). To efficiently find the GMEC, protein design algorithms must methodically reduce the conformational search space. By applying distance and energy cutoffs, the protein system to be designed can thus be represented using a sparse residue interaction graph, where the number of interacting residue pairs is less than all pairs of mutable residues, and the corresponding GMEC is called the sparse GMEC. However, ignoring some pairwise residue interactions can lead to a change in the energy, conformation, or sequence of the sparse GMEC vs. the original or the full GMEC. Despite the widespread use of sparse residue interaction graphs in protein design, the above mentioned effects of their use have not been previously analyzed. To analyze the costs and benefits of designing with sparse residue interaction graphs, we computed the GMECs for 136 different protein design problems both with and without distance and energy cutoffs, and compared their energies, conformations, and sequences. Our analysis shows that the differences between the GMECs depend critically on whether or not the design includes core, boundary, or surface residues. Moreover, neglecting long-range interactions can alter local interactions and introduce large sequence differences, both of which can result in significant structural and functional changes. Designs on proteins with experimentally measured thermostability show it is beneficial to compute both the full and the sparse GMEC accurately and efficiently. To this end, we show that a provable, ensemble-based algorithm can efficiently compute both GMECs by enumerating a small number of conformations, usually fewer than 1000. This provides a novel way to combine sparse residue interaction graphs with provable, ensemble-based algorithms to reap the benefits of sparse residue interaction graphs while avoiding their potential inaccuracies.

Computational structure-based protein design algorithms have successfully redesigned proteins to fold and bind target substrates in vitro, and even in vivo. Because the complexity of a computational design increases dramatically with the number of mutable residues, many design algorithms employ cutoffs (distance or energy) to neglect some pairwise residue interactions, thereby reducing the effective search space and computational cost. However, the energies neglected by such cutoffs can add up, which may have nontrivial effects on the designed sequence and its function. To study the effects of using cutoffs on protein design, we computed the optimal sequence both with and without cutoffs, and showed that neglecting long-range interactions can significantly change the computed conformation and sequence. Designs on proteins with experimentally measured thermostability showed the benefits of computing the optimal sequences (and their conformations), both with and without cutoffs, efficiently and accurately. Therefore, we also showed that a provable, ensemble-based algorithm can efficiently compute the optimal conformation and sequence, both with and without applying cutoffs, by enumerating a small number of conformations, usually fewer than 1000. This provides a novel way to combine cutoffs with provable, ensemble-based algorithms to reap the computational efficiency of cutoffs while avoiding their potential inaccuracies.

Generative models of conformational dynamics

Atomistic simulations of the conformational dynamics of proteins can be performed using either Molecular Dynamics or Monte Carlo procedures. The ensembles of three-dimensional structures produced during simulation can be analyzed in a number of ways to elucidate the thermodynamic and kinetic properties of the system. The goal of this chapter is to review both traditional and emerging methods for learning generative models from atomistic simulation data. Here, the term 'generative' refers to a model of the joint probability distribution over the behaviors of the constituent atoms. In the context of molecular modeling, generative models reveal the correlation structure between the atoms, and may be used to predict how the system will respond to structural perturbations. We begin by discussing traditional methods, which produce multivariate Gaussian models. We then discuss GAMELAN (GRAPHICAL MODELS OF ENERGY LANDSCAPES), which produces generative models of complex, non-Gaussian conformational dynamics (e.g., allostery, binding, folding, etc.) from long timescale simulation data.

Computational design of a PDZ domain peptide inhibitor that rescues CFTR activity

The cystic fibrosis transmembrane conductance regulator (CFTR) is an epithelial chloride channel mutated in patients with cystic fibrosis (CF). The most prevalent CFTR mutation, ΔF508, blocks folding in the endoplasmic reticulum. Recent work has shown that some ΔF508-CFTR channel activity can be recovered by pharmaceutical modulators (“potentiators” and “correctors”), but ΔF508-CFTR can still be rapidly degraded via a lysosomal pathway involving the CFTR-associated ligand (CAL), which binds CFTR via a PDZ interaction domain. We present a study that goes from theory, to new structure-based computational design algorithms, to computational predictions, to biochemical testing and ultimately to epithelial-cell validation of novel, effective CAL PDZ inhibitors (called “stabilizers”) that rescue ΔF508-CFTR activity. To design the “stabilizers”, we extended our structural ensemble-based computational protein redesign algorithm to encompass protein-protein and protein-peptide interactions. The computational predictions achieved high accuracy: all of the top-predicted peptide inhibitors bound well to CAL. Furthermore, when compared to state-of-the-art CAL inhibitors, our design methodology achieved higher affinity and increased binding efficiency. The designed inhibitor with the highest affinity for CAL (kCAL01) binds six-fold more tightly than the previous best hexamer (iCAL35), and 170-fold more tightly than the CFTR C-terminus. We show that kCAL01 has physiological activity and can rescue chloride efflux in CF patient-derived airway epithelial cells. Since stabilizers address a different cellular CF defect from potentiators and correctors, our inhibitors provide an additional therapeutic pathway that can be used in conjunction with current methods.

Cystic fibrosis (CF) is an inherited disease that causes the body to produce thick mucus that clogs the lungs and obstructs the breakdown and absorption of food. The cystic fibrosis transmembrane conductance regulator (CFTR) is mutated in CF patients, and the most common mutation causes three defects in CFTR: misfolding, decreased function, and rapid degradation. Drugs are currently being studied to correct the first two CFTR defects, but the problem of rapid degradation remains. Recently, key protein-protein interactions have been discovered that implicate the protein CAL in CFTR degradation. Here we have developed new computational protein design algorithms and used them to successfully predict peptide inhibitors of the CAL-CFTR interface. Our algorithm uses a structural ensemble-based evaluation of protein sequences and conformations to calculate accurate predictions of protein-peptide binding affinities. The algorithm is general and can be applied to a wide variety of protein-protein interface designs. All of our designed inhibitors bound CAL with high affinity. We tested our top binding peptide and observed that the inhibitor could successfully rescue CFTR function in CF patient-derived epithelial cells. Our designed inhibitors provide a novel therapeutic path which could be used in combination with existing CF therapeutics for additive benefit.

Exposing the co-adaptive potential of protein-protein interfaces through computational sequence design

MOTIVATION

In nature, protein-protein interactions are constantly evolving under various selective pressures. Nonetheless, it is expected that crucial interactions are maintained through compensatory mutations between interacting proteins. Thus, many studies have used evolutionary sequence data to extract such occurrences of correlated mutation. However, this research is confounded by other evolutionary pressures that contribute to sequence covariance, such as common ancestry.

RESULTS

Here, we focus exclusively on the compensatory mutations deriving from physical protein interactions, by performing large-scale computational mutagenesis experiments for >260 protein-protein interfaces. We investigate the potential for co-adaptability present in protein pairs that are always found together in nature (obligate) and those that are occasionally in complex (transient). By modeling each complex both in bound and unbound forms, we find that naturally transient complexes possess greater relative capacity for correlated mutation than obligate complexes, even when differences in interface size are taken into account.

Design of multispecific protein sequences using probabilistic graphical modeling

In nature, proteins partake in numerous protein- protein interactions that mediate their functions. Moreover, proteins have been shown to be physically stable in multiple structures, induced by cellular conditions, small ligands, or covalent modifications. Understanding how protein sequences achieve this structural promiscuity at the atomic level is a fundamental step in the drug design pipeline and a critical question in protein physics. One way to investigate this subject is to computationally predict protein sequences that are compatible with multiple states, i.e., multiple target structures or binding to distinct partners. The goal of engineering such proteins has been termed multispecific protein design. We develop a novel computational framework to efficiently and accurately perform multispecific protein design. This framework utilizes recent advances in probabilistic graphical modeling to predict sequences with low energies in multiple target states. Furthermore, it is also geared to specifically yield positional amino acid probability profiles compatible with these target states. Such profiles can be used as input to randomly bias high-throughput experimental sequence screening techniques, such as phage display, thus providing an alternative avenue for elucidating the multispecificity of natural proteins and the synthesis of novel proteins with specific functionalities. We prove the utility of such multispecific design techniques in better recovering amino acid sequence diversities similar to those resulting from millions of years of evolution. We then compare the approaches of prediction of low energy ensembles and of amino acid profiles and demonstrate their complementarity in providing more robust predictions for protein design.

Tradeoff between stability and multispecificity in the design of promiscuous proteins

Natural proteins often partake in several highly specific protein-protein interactions. They are thus subject to multiple opposing forces during evolutionary selection. To be functional, such multispecific proteins need to be stable in complex with each interaction partner, and, at the same time, to maintain affinity toward all partners. How is this multispecificity acquired through natural evolution? To answer this compelling question, we study a prototypical multispecific protein, calmodulin (CaM), which has evolved to interact with hundreds of target proteins. Starting from high-resolution structures of sixteen CaM-target complexes, we employ state-of-the-art computational methods to predict a hundred CaM sequences best suited for interaction with each individual CaM target. Then, we design CaM sequences most compatible with each possible combination of two, three, and all sixteen targets simultaneously, producing almost 70,000 low energy CaM sequences. By comparing these sequences and their energies, we gain insight into how nature has managed to find the compromise between the need for favorable interaction energies and the need for multispecificity. We observe that designing for more partners simultaneously yields CaM sequences that better match natural sequence profiles, thus emphasizing the importance of such strategies in nature. Furthermore, we show that the CaM binding interface can be nicely partitioned into positions that are critical for the affinity of all CaM-target complexes and those that are molded to provide interaction specificity. We reveal several basic categories of sequence-level tradeoffs that enable the compromise necessary for the promiscuity of this protein. We also thoroughly quantify the tradeoff between interaction energetics and multispecificity and find that facilitating seemingly competing interactions requires only a small deviation from optimal energies. We conclude that multispecific proteins have been subjected to a rigorous optimization process that has fine-tuned their sequences for interactions with a precise set of targets, thus conferring their multiple cellular functions.

In nature, some proteins are more social than others, interacting with a large number of partners. These “promiscuous” proteins play key roles in cellular signaling pathways whose disruption may lead to diseases such as cancer. The amino acid sequences of such proteins must have evolved to be optimal for combined interactions with all natural partners. However, the evolutionary process leading to this promiscuity is not fully understood. We address this subject by predicting amino acid sequences that would be most compatible for interaction with each partner on its own and those most compatible for binding multiple proteins. We find that these two types of sequences are substantially different, the latter more closely resembling the natural sequences of promiscuous proteins. We also find that promiscuous proteins contain certain regions that are necessary for interfacing with all of their partners, while other regions convey specific interactions with each particular target protein. We analyze the tradeoffs required for such proteins to bind multiple partners and find that only some degree of compromise is typically needed in order to permit interactions that are seemingly antagonistic. We conclude that the simulations reported here mimic well the natural evolution of proteins that associate with multiple partners.

Accurate prediction for atomic-level protein design and its application in diversifying the near-optimal sequence space

The task of engineering a protein to assume a target three-dimensional structure is known as protein design. Computational search algorithms are devised to predict a minimal energy amino acid sequence for a particular structure. In practice, however, an ensemble of low-energy sequences is often sought. Primarily, this is performed because an individual predicted low-energy sequence may not necessarily fold to the target structure because of both inaccuracies in modeling protein energetics and the nonoptimal nature of search algorithms employed. Additionally, some low-energy sequences may be overly stable and thus lack the dynamic flexibility required for biological functionality. Furthermore, the investigation of low-energy sequence ensembles will provide crucial insights into the pseudo-physical energy force fields that have been derived to describe structural energetics for protein design. Significantly, numerous studies have predicted low-energy sequences, which were subsequently synthesized and demonstrated to fold to desired structures. However, the characterization of the sequence space defined by such energy functions as compatible with a target structure has not been performed in full detail. This issue is critical for protein design scientists to successfully continue using these force fields at an ever-increasing pace and scale. In this paper, we present a conceptually novel algorithm that rapidly predicts the set of lowest energy sequences for a given structure. Based on the theory of probabilistic graphical models, it performs efficient inspection and partitioning of the near-optimal sequence space, without making any assumptions of positional independence. We benchmark its performance on a diverse set of relevant protein design examples and show that it consistently yields sequences of lower energy than those derived from state-of-the-art techniques. Thus, we find that previously presented search techniques do not fully depict the low-energy space as precisely. Examination of the predicted ensembles indicates that, for each structure, the amino acid identity at a majority of positions must be chosen extremely selectively so as to not incur significant energetic penalties. We investigate this high degree of similarity and demonstrate how more diverse near-optimal sequences can be predicted in order to systematically overcome this bottleneck for computational design. Finally, we exploit this in-depth analysis of a collection of the lowest energy sequences to suggest an explanation for previously observed experimental design results. The novel methodologies introduced here accurately portray the sequence space compatible with a protein structure and further supply a scheme to yield heterogeneous low-energy sequences, thus providing a powerful instrument for future work on protein design.