De novo protein design by an energy function based on series expansion in distance and orientation dependence

ProBID-Net: a deep learning model for protein-protein binding interface design

Protein-protein interactions are pivotal in numerous biological processes. The computational design of these interactions facilitates the creation of novel binding proteins, crucial for advancing biopharmaceutical products. With the evolution of artificial intelligence (AI), protein design tools have swiftly transitioned from scoring-function-based to AI-based models. However, many AI models for protein design are constrained by assuming complete unfamiliarity with the amino acid sequence of the input protein, a feature most suited for de novo design but posing challenges in designing protein-protein interactions when the receptor sequence is known. To bridge this gap in computational protein design, we introduce ProBID-Net. Trained using natural protein-protein complex structures and protein domain-domain interface structures, ProBID-Net can discern features from known target protein structures to design specific binding proteins based on their binding sites. In independent tests, ProBID-Net achieved interface sequence recovery rates of 52.7%, 43.9%, and 37.6%, surpassing or being on par with ProteinMPNN in binding protein design. Validated using AlphaFold-Multimer, the sequences designed by ProBID-Net demonstrated a close correspondence between the design target and the predicted structure. Moreover, the model's output can predict changes in binding affinity upon mutations in protein complexes, even in scenarios where no data on such mutations were provided during training (zero-shot prediction). In summary, the ProBID-Net model is poised to significantly advance the design of protein-protein interactions.

All-Atom Protein Sequence Design Based on Geometric Deep Learning

Designing sequences for specific protein backbones is a key step in creating new functional proteins. Here, we introduce GeoSeqBuilder, a deep learning framework that integrates protein sequence generation with side chain conformation prediction to produce the complete all-atom structures for designed sequences. GeoSeqBuilder uses spatial geometric features from protein backbones and explicitly includes three-body interactions of neighboring residues. GeoSeqBuilder achieves native residue type recovery rate of 51.6 %, comparable to ProteinMPNN and other leading methods, while accurately predicting side chain conformations. We first used GeoSeqBuilder to design sequences for thioredoxin and a hallucinated three-helical bundle protein. All the 15 tested sequences expressed as soluble monomeric proteins with high thermal stability, and the 2 high-resolution crystal structures solved closely match the designed models. The generated protein sequences exhibit low similarity (minimum 23 %) to the original sequences, with significantly altered hydrophobic cores. We further redesigned the hydrophobic core of glutathione peroxidase 4, and 3 of the 5 designs showed improved enzyme activity. Although further testing is needed, the high experimental success rate in our testing demonstrates that GeoSeqBuilder is a powerful tool for designing novel sequences for predefined protein structures with atomic details. GeoSeqBuilder is available at https://github.com/PKUliujl/GeoSeqBuilder.

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.

Ab Initio Prediction of 3-D Conformations for Protein Long Loops with High Accuracy and Applications to Antibody CDRH3 Modeling

Residue-level potentials of mean force were widely used for protein backbone refinements to avoid simultaneous sampling of side-chain conformations. The interaction energy between the reduced side chains and backbone atoms was not considered explicitly. In this study, we developed novel methods to calculate the residue-atom interaction energy in combination with atomic and residue-level terms. The parameters were optimized step by step to remove the overcounting or overlap problem between different energy terms. The mixing energy functions were then used to evaluate the generated backbone conformations at the initial sampling stage of protein loop modeling (OSCAR-loop), including the interaction energy between the reduced loop residues and full atoms of the protein framework. The accuracies of top-ranked decoys were 1.18 and 2.81 Å for 8-residue and 12-residue loops, respectively. We then selected diverse decoys for side-chain modeling, backbone refinement, and energy minimization. The procedure was repeated multiple times to select one prediction with the lowest energy. Consequently, we obtained an accuracy of 0.74 Å for a prevailing test set of 12-residue loops, compared with >1.4 Å reported by other researchers. The OSCAR-loop was also effective for modeling the H3 loops of antibody complementary determining regions (CDRs) in the crystal environment. The prediction accuracy of OSCAR-loop (1.74 Å) was better than the accuracy of the Rosetta NGK method (3.11 Å) or those achieved by deep learning methods (>2.2 Å) for the CDRH3 loops of 49 targets in the Rosetta antibody benchmark. The performance of OSCAR-loop in a model environment was also discussed.

SPIN-CGNN: Improved fixed backbone protein design with contact map-based graph construction and contact graph neural network

Recent advances in deep learning have significantly improved the ability to infer protein sequences directly from protein structures for the fix-backbone design. The methods have evolved from the early use of multi-layer perceptrons to convolutional neural networks, transformers, and graph neural networks (GNN). However, the conventional approach of constructing K-nearest-neighbors (KNN) graph for GNN has limited the utilization of edge information, which plays a critical role in network performance. Here we introduced SPIN-CGNN based on protein contact maps for nearest neighbors. Together with auxiliary edge updates and selective kernels, we found that SPIN-CGNN provided a comparable performance in refolding ability by AlphaFold2 to the current state-of-the-art techniques but a significant improvement over them in term of sequence recovery, perplexity, deviation from amino-acid compositions of native sequences, conservation of hydrophobic positions, and low complexity regions, according to the test by unseen structures, "hallucinated" structures and diffusion models. Results suggest that low complexity regions in the sequences designed by deep learning, for generated structures in particular, remain to be improved, when compared to the native sequences.

ZetaDesign: an end-to-end deep learning method for protein sequence design and side-chain packing

Computational protein design has been demonstrated to be the most powerful tool in the last few years among protein designing and repacking tasks. In practice, these two tasks are strongly related but often treated separately. Besides, state-of-the-art deep-learning-based methods cannot provide interpretability from an energy perspective, affecting the accuracy of the design. Here we propose a new systematic approach, including both a posterior probability and a joint probability parts, to solve the two essential questions once for all. This approach takes the physicochemical property of amino acids into consideration and uses the joint probability model to ensure the convergence between structure and amino acid type. Our results demonstrated that this method could generate feasible, high-confidence sequences with low-energy side conformations. The designed sequences can fold into target structures with high confidence and maintain relatively stable biochemical properties. The side chain conformation has a significantly lower energy landscape without delegating to a rotamer library or performing the expensive conformational searches. Overall, we propose an end-to-end method that combines the advantages of both deep learning and energy-based methods. The design results of this model demonstrate high efficiency, and precision, as well as a low energy state and good interpretability.

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.