This study integrates large language model-guided antibody design with experimental validation to assess the real-world performance of computationally optimized antibodies, and will clarify how effectively AI-generated sequences translate into functional high-affinity binders, informing the development of future therapeutics.
Abstract
Machine-learning—guided antibody design offers a promising approach for rapidly generating high-affinity therapeutics against emerging pathogens. Our group has developed Ab-Affinity, a large language model that predicts antibody—antigen binding and, together with genetic algorithms and simulated annealing, designs variants with markedly improved predicted stability and affinity for a SARS-CoV-2 spike epitope. Computational analyses indicate over a 160-fold affinity enhancement compared to experimentally derived sequences. This work focuses on experimentally validating these predictions through expression, purification, and functional characterization of the model-designed antibodies
Three antibodies–Ab-14-seed and the optimized variants Ab-14-SA-PSSM1 and Ab-14-SA-PSSM6–designed through Ab-Affinity were expressed in Pichia pastoris and purified using FPLC. Protein expression and purity were confirmed by SDS-PAGE and Western blot. Binding affinities are being assessed using ELISA and surface plasmon resonance to evaluate interactions with the target SARS-CoV-2 spike peptide.
All three designed antibodies have been successfully expressed in Pichia pastoris and purified. Upcoming binding affinity assays will determine whether these variants demonstrate the enhanced binding affinities predicted by Ab-Affinity.
This study integrates large language model-guided antibody design with experimental validation to assess the real-world performance of computationally optimized antibodies. The results will clarify how effectively AI-generated sequences translate into functional high-affinity binders, informing the development of future therapeutics.
DOE
Vaccines and Immunotherapy (VAC)
The AIntibody challenge shows that AI can optimize antibodies in defined, biologically grounded regimes, in addition to highlighting critical gaps including affinity prediction and library-inspired antibody design and cross-task generalization.
M. Erasmus, Daniel Bedinger, Elizabeth Hopkins et al.· Nature Biotechnology· 0 citations
Antibodies raised against human targets often fail to recognize their animal orthologs, limiting preclinical evaluation in relevant models. We developed a Deep Mutational Scanning (DMS)-coupled deep learning strategy to engineer potent cross-reactive antibodies with minimal sequence divergence. Starting from C4, a fully human anti-PD-L1 antibody with weak recognition of murine PD-L1, DMS identified substitutions that improved binding to both human and mouse antigens. Conventional recombination of beneficial mutations generated highly cross-reactive antibodies but required 13 to 15 substitutions. To reduce this mutational burden, a deep learning model trained on DMS-derived sequence-binding data was used to identify minimal mutation combinations predicted to retain high affinity. This approach yielded variants carrying only 4 to 5 substitutions, with in vitro and cellular binding properties comparable to highly mutated antibodies. Epitope mapping, structural modeling and in vivo assessment further confirmed that these engineered antibodies retained PD-1/PD-L1 blockade and demonstrated therapeutic activity in a mouse tumor model.
H. Dorison, Anne-Laure Grindel, François Thenier et al.· bioRxiv· 0 citations
These findings provide practical guidance for integrating open-source protein structure prediction models into AI-driven nanobody discovery pipelines while highlighting the need for improved generalization across antigens.
Yannick Vogt, Rebekka Roßberg, Jan Habermann et al.· Frontiers in Bioinformatics· 1 citation
Functional cure of chronic hepatitis B virus (HBV) infection remains a significant challenge, making viral-entry-blocking antibodies a promising antiviral strategy. Here, we developed a structure-guided computational workflow for affinity-enhancing candidate mutations at the interface between the humanized neutralizing antibody HzKR127 and the HBV preS1 peptide epitope. Based on the crystal structure of the HzKR127–preS1 complex, we performed single-site saturation mutagenesis across the paratope, evaluating variants with a consensus effect score integrated from seven computational models. Benchmarking against published alanine scanning data showed that our consensus score effectively identified major-affinity-loss residues, achieving ROC AUC values of 0.81 and 0.84 for residue-level and site-level predictions, respectively. Mutational profiling revealed distinct asymmetric mutational responses, with broad intolerance on the preS1 side and localized favorable substitutions within antibody CDRs. Multilevel prioritization identified 26 antibody-side candidates, 17 of which showed improved HADDOCK refinement scores compared to the wild type. In particular, the H:D97W/F/Y substitutions presented the strongest structural rationale for enhancing improved interfacial packing through aromatic hydrophobic contacts with preS1 Phe10. These findings provide a prioritized list of candidates for experimental validation and a practical framework for the rational optimization of antibodies targeting functionally constrained viral epitopes.
Wen-Qing Chen, Yuan-Zhong Tu, Kai Wang et al.· Current Issues in Molecular...· 0 citations
DyAb is a pair-wise representation built on top of a pre-trained protein language model that achieves a Spearman rank correlation of up to 0.85 on binding affinity prediction across monoclonal antibodies targeting three different antigens.
J. Lin, Jennifer L. Hofmann, Andrew Leaver-Fay et al.· mAbs· 0 citations
Existing immunological datasets can support feature-constrained transfer learning for data-efficient prioritization of antibody:antigen interactions across closely related Sarbecoviruses, particularly when conserved epitope regions can be aligned and limited target-specific measurements are available for calibration.
Jimmy Yuan, Ryan Bruneau, Stephen Won et al.· Frontiers in Immunology· 0 citations
