De novo sequence redesign of a functional Ras‐binding domain globally inverted the surface charge distribution and led to extreme thermostability

Abstract

To acquire extremely thermostable proteins of given functions is challenging for conventional protein engineering. Here we applied ABACUS, a statistical energy function we developed for de novo amino acid sequence design, to globally redesign a Ras‐binding domain (RBD), and obtained an extremely thermostable RBD that unfolds reversibly at above 110°C, the redesigned RBD experimentally confirmed to have expected structure and Ras‐binding interface. Directed evolution of the redesigned RBD improved its Ras‐binding affinity to the native protein level without excessive loss of thermostability. The designed amino acid substitutions were mostly at the protein surface. For many substitutions, strong epistasis or significantly differentiated effects on thermostability in the native sequence context relative to the redesigned sequence context were observed, suggesting the globally redesigned sequence to be unreachable through combining beneficial mutations of the native sequence. Further analyses revealed that by replacing 38 of a total of 48 non‐interfacial surface residues at once, ABACUS redesign was able to globally “invert” the protein's charge distribution pattern in an optimized way. Our study demonstrates that computational protein design provides powerful new tools to solve challenging protein engineering problems.

Rotamer-free protein sequence design based on deep learning and self-consistency

“…Previously, we developed a statistical energy model named ABACUS [9,10] (a backbone-based amino acid usage survey). Proteins designed with ABACUS have been verified with experimentally-solved structures in a number of studies [9,10,22,23,24]. Although the energy terms in ABACUS were devised to consider high-order coupling be-tween various physical factors, the non-linear integration of coupled effects was restricted to the single-residue and the residue-pair-wise levels, beyond which separately learned energy terms were combined only in a linear way.…”

Computer-Aided Drug Design Boosts RAS Inhibitor Discovery

“…Recently, proteins such as fluorescence-activating β-barrel were developed by de novo drug design [ 137 ]. A functional RAS-binding domain with extreme thermostability was identified by another de novo sequence redesign model called ABACUS by Liu et al [ 138 ]. The de novo sequence redesign model does not suffer from a restrictively cumulative effect for future directions.…”

Computational protein design with data‐driven approaches: Recent developments and perspectives

“…The use of jointly conditioned sidechain preferences to estimate the statistical energies led to a model that was experimentally verified 59 to have comparative design success rates and complementary design results as state-of-the-art physics-based energy models, or statistical energy models fitted with continuous structural variables. 51,[60][61][62] Successful redesigns by ABACUS on natural backbones led to proteins of very high thermal stability (with unfolding temperatures approaching or going above 100 C), 63,64 sometimes through the automatically re-optimized electrostatic interaction network at protein surfaces. 64 These observations suggest that with currently available protein data, appropriate data-driven models are able to comprehensively recapitulate the intrinsic biophysical principles underlying the folded protein structures.…”

“…51,[60][61][62] Successful redesigns by ABACUS on natural backbones led to proteins of very high thermal stability (with unfolding temperatures approaching or going above 100 C), 63,64 sometimes through the automatically re-optimized electrostatic interaction network at protein surfaces. 64 These observations suggest that with currently available protein data, appropriate data-driven models are able to comprehensively recapitulate the intrinsic biophysical principles underlying the folded protein structures.…”