Generation of bright monomeric red fluorescent proteins via computational design of enhanced chromophore packing

Red fluorescent proteins (RFPs) have found widespread application in chemical and biological research due to their longer emission wavelengths. Here, we use computational protein design to increase the quantum yield...

Protein Design with Deep Learning

Computational Protein Design (CPD) has produced impressive results for engineering new proteins, resulting in a wide variety of applications. In the past few years, various efforts have aimed at replacing or improving existing design methods using Deep Learning technology to leverage the amount of publicly available protein data. Deep Learning (DL) is a very powerful tool to extract patterns from raw data, provided that data are formatted as mathematical objects and the architecture processing them is well suited to the targeted problem. In the case of protein data, specific representations are needed for both the amino acid sequence and the protein structure in order to capture respectively 1D and 3D information. As no consensus has been reached about the most suitable representations, this review describes the representations used so far, discusses their strengths and weaknesses, and details their associated DL architecture for design and related tasks.

Computational and experimental assessment of backbone templates for computational protein design

Computational protein design has taken big strides over the recent years, however, the tools available are still not at a state where a sequence can be designed to fold into a given protein structure at will and with high probability. We have here applied a recent release of Rosetta Design to redesign a set of structurally very similar proteins belonging to the Thioredoxin fold. We determined design success using a combination of a genetic screening tool to assay folding/stability in E. coli and selecting the best hits from this for further biochemical characterization. We have previously used this set of template proteins for redesign and found that success was highly dependent on template structure, a trait which was also found in this study. Nevertheless, state of the art design software is now able to predict the best template, most likely due to the introduction of the cart_bonded energy term. The template that led to the greatest fraction of successful designs was the same (a Thioredoxin from spinach) as that identified in our previous study. Our previously described redesign of Thioredoxin, which also used the spinach protein as template, however also performed well. In the present study, both these templates yielded proteins with compact folded structures, and enforces the conclusion that any design project must carefully consider different design templates. Fortunately, selecting designs using the cart_bonded energy term appears to correctly identify such templates.

Guaranteed Diversity and Optimality in Cost Function Network Based Computational Protein Design Methods

Proteins are the main active molecules of life. Although natural proteins play many roles, as enzymes or antibodies for example, there is a need to go beyond the repertoire of natural proteins to produce engineered proteins that precisely meet application requirements, in terms of function, stability, activity or other protein capacities. Computational Protein Design aims at designing new proteins from first principles, using full-atom molecular models. However, the size and complexity of proteins require approximations to make them amenable to energetic optimization queries. These approximations make the design process less reliable, and a provable optimal solution may fail. In practice, expensive libraries of solutions are therefore generated and tested. In this paper, we explore the idea of generating libraries of provably diverse low-energy solutions by extending cost function network algorithms with dedicated automaton-based diversity constraints on a large set of realistic full protein redesign problems. We observe that it is possible to generate provably diverse libraries in reasonable time and that the produced libraries do enhance the Native Sequence Recovery, a traditional measure of design methods reliability.

Guaranted Diversity and Optimality in Cost Function Network Based Computational Protein Design Methods

Proteins are the main active molecules of Life. While natural proteins play many roles, as enzymes or antibodies for example, there is a need to go beyond the repertoire of natural proteins to produce engineered proteins that precisely meet application requirements, in terms of function, stability, activity or other protein capacities. Computational Protein Design aims at designing new proteins from first principles, using full-atom molecular models. However, the size and complexity of proteins require approximations to make them amenable to energetic optimization queries. These approximations make the design process less reliable and a provable optimal solution may fail. In practice, expensive libraries of solutions are therefore generated and tested. In this paper, we explore the idea of generating libraries of provably diverse low energy solutions by extending Cost Function Network algorithms with dedicated automaton-based diversity constraints on a large set of realistic full protein redesign problems. We observe that it is possible to generate provably diverse libraries in reasonable time and that the produced libraries do enhance the Native Sequence Recovery, a traditional measure of design methods reliability.

The Role of Rigid Residues in Modulating TEM-1 β-Lactamase Function and Thermostability

The relationship between protein motions (i.e., dynamics) and enzymatic function has begun to be explored in β-lactamases as a way to advance our understanding of these proteins. In a recent study, we analyzed the dynamic profiles of TEM-1 (a ubiquitous class A β-lactamase) and several ancestrally reconstructed homologues. A chief finding of this work was that rigid residues that were allosterically coupled to the active site appeared to have profound effects on enzyme function, even when separated from the active site by many angstroms. In the present work, our aim was to further explore the implications of protein dynamics on β-lactamase function by altering the dynamic profile of TEM-1 using computational protein design methods. The Rosetta software suite was used to mutate amino acids surrounding either rigid residues that are highly coupled to the active site or to flexible residues with no apparent communication with the active site. Experimental characterization of ten designed proteins indicated that alteration of residues surrounding rigid, highly coupled residues, substantially affected both enzymatic activity and stability; in contrast, native-like activities and stabilities were maintained when flexible, uncoupled residues, were targeted. Our results provide additional insight into the structure-function relationship present in the TEM family of β-lactamases. Furthermore, the integration of computational protein design methods with analyses of protein dynamics represents a general approach that could be used to extend our understanding of the relationship between dynamics and function in other enzyme classes.

Iterated Local Search with Partition Crossover for Computational Protein Design

Structure-based computational protein design (CPD) refers to the problem of finding a sequence of amino acids which folds into a specific desired protein structure, and possibly fulfills some targeted biochemical properties. Recent studies point out the particularly rugged CPD energy landscape, suggesting that local search optimization methods should be designed and tuned to easily escape local minima attraction basins. In this paper, we analyze the performance and search dynamics of an iterated local search (ILS) algorithm enhanced with partition crossover. Our algorithm, PILS, quickly finds local minima and escapes their basins of attraction by solution perturbation. Additionally, the partition crossover operator exploits the structure of the residue interaction graph in order to efficiently mix solutions and find new unexplored basins. Our results on a benchmark of 30 proteins of various topology and size show that PILS consistently finds lower energy solutions compared to Rosetta fixbb and a classic ILS, and that the corresponding sequences are mostly closer to the native.