Computational and Experimental Assessment of Backbone Templates for Computational Redesign of the Thioredoxin Fold

Protein sequence design on given backbones with deep learning

Deep learning methods for protein sequence design focus on modeling and sampling the many- dimensional distribution of amino acid sequences conditioned on the backbone structure. To produce physically foldable sequences, inter-residue couplings need to be considered properly. These couplings are treated explicitly in iterative methods or autoregressive methods. Non-autoregressive models treating these couplings implicitly are computationally more efficient, but still await tests by wet experiment. Currently, sequence design methods are evaluated mainly using native sequence recovery rate and native sequence perplexity. These metrics can be complemented by sequence-structure compatibility metrics obtained from energy calculation or structure prediction. However, existing computational metrics have important limitations that may render the generalization of computational test results to performance in real applications unwarranted. Validation of design methods by wet experiments should be encouraged.

Dissecting the stability determinants of a challenging de novo protein fold using massively parallel design and experimentation

Designing entirely new protein structures remains challenging because we do not fully understand the biophysical determinants of folding stability. Yet, some protein folds are easier to design than others. Previous work identified the 43-residue ɑββɑ fold as especially challenging: The best designs had only a 2% success rate, compared to 39 to 87% success for other simple folds [G. J. Rocklin et al., Science 357, 168-175 (2017)]. This suggested the ɑββɑ fold would be a useful model system for gaining a deeper understanding of folding stability determinants and for testing new protein design methods. Here, we designed over 10,000 new ɑββɑ proteins and found over 3,000 of them to fold into stable structures using a high-throughput protease-based assay. NMR, hydrogen-deuterium exchange, circular dichroism, deep mutational scanning, and scrambled sequence control experiments indicated that our stable designs fold into their designed ɑββɑ structures with exceptional stability for their small size. Our large dataset enabled us to quantify the influence of universal stability determinants including nonpolar burial, helix capping, and buried unsatisfied polar atoms, as well as stability determinants unique to the ɑββɑ topology. Our work demonstrates how large-scale design and test cycles can solve challenging design problems while illuminating the biophysical determinants of folding.

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

Several previously proposed deep learning methods to design amino acid sequences that autonomously fold into a given protein backbone yielded promising results in computational tests but did not outperform conventional energy function-based methods in wet experiments. Here we present the ABACUS-R method, which uses an encoder-decoder network trained using a multitask learning strategy to predict the sidechain type of a central residue from its three-dimensional local environment, which includes, besides other features, the types but not the conformations of the surrounding sidechains. This eliminates the need to reconstruct and optimize sidechain structures, and drastically simplifies the sequence design process. Thus iteratively applying the encoder-decoder to different central residues is able to produce self-consistent overall sequences for a target backbone. Results of wet experiments, including five structures solved by X-ray crystallography, show that ABACUS-R outperforms state-of-the-art energy function-based methods in success rate and design precision.