Abstract
Generative artificial intelligence (AI) has the potential to greatly increase the speed, quality and controllability of antibody design. Traditional de novo antibody discovery requires time and resource intensive screening of large immune or synthetic libraries. These methods also offer little control over the output sequences, which can result in lead candidates with sub-optimal binding and poor developability attributes. Several groups have introduced models for generative antibody design with promising in silico evidence [1–10], however, no such method has demonstrated generative AI-based de novo antibody design with experimental validation. Here we use generative deep learning models to de novo design antibodies against three distinct targets, in a zero-shot fashion, where all designs are the result of a single round of model generations with no follow-up optimization. In particular, we screen over 1 million antibody variants designed for binding to human epidermal growth factor receptor 2 (HER2) using our high-throughput wet lab capabilities. Our models successfully design all CDRs in the heavy chain of the antibody and compute likelihoods that are calibrated with binding. We achieve binding rates of 10.6% and 1.8% for heavy chain CDR3 (HCDR3) and HCDR123 designs respectively, which is four and eleven times higher than HCDR3s and HCDR123s randomly sampled from the Observed Antibody Space (OAS) [11]. We further characterize 421 AI-designed binders using surface plasmon resonance (SPR), finding three that bind tighter than the therapeutic antibody trastuzumab. The binders are highly diverse, have low sequence identity to known antibodies, and adopt variable structural conformations. Additionally, the binders score highly on our previously introduced Naturalness metric [12], indicating they are likely to possess desirable developability profiles and low immunogenicity. We open source1 the HER2 binders and report the measured binding affinities. These results unlock a path to accelerated drug creation for novel therapeutic targets using generative AI and high-throughput experimentation.
Competing Interest Statement
The authors are current or former employees, contractors, interns, or executives of Absci Corporation and may hold shares in Absci Corporation. Methods and compositions described in this manuscript are the subject of one or more pending patent applications.
Footnotes
↵* Equal contribution
*Extensive characterization of model hit rate (>10% for HCDR123) * Achievement of zero-shot HCDR123 design * Comparisons to biological baselines: models outperform random OAS by >10x in HCDR123 setting * Description of fast experimental cycle times * Author list updated to reflect new contributions
https://github.com/AbSciBio/unlocking-de-novo-antibody-design
↵1 https://github.com/AbsciBio/unlocking-de-novo-antibody-design
↵3 Motivation is given in the official Rosetta documentation for Fast Relax