Absci’s ambition to grow into the Google search engine of protein-based drug discovery and biomanufacturing reached a milestone this month when the company announced that its artificial intelligence (AI) platform has succeeded in creating and validating de novo antibodies in silico.
The milestone, Absci asserts, may finally deliver what drug developers have long promised, namely the ability to slash the time it takes to develop new drugs, and thus lower the cost of developing treatments as they reach the market—a reduction the company says could also lead to lower-cost treatments for patients. Absci says its AI-designed antibodies can cut drug discovery timeframes by more than 50%, from as much as six years down to 18-24 months, while also increasing their probability of success in the clinic.
Absci said it achieved its results through its zero-shot generative AI method, which designs antibodies to bind to specific targets without incorporating any training data on antibodies that are known to bind those targets, forcing the model to design the antibody from scratch. Absci said its model produced antibody designs that were unlike those found in existing antibody databases, and which worked in lab tests without the need for first optimizing the in silico designs.
The company’s wet lab validated the superiority of its de novo antibody candidates to bind to human epidermal growth factor receptor 2 (HER2) and three additional targets—human vascular endothelial growth factor A (VEGF-A), the Omicron variant of SARS-CoV-2 spike receptor-binding domain (COVID-19), and a third, undisclosed target.
“We didn’t want to just show that it could work with just one, but the technology was broadly applicable to really any type of target you want it to work on,” Absci Founder and CEO Sean McClain told GEN Edge.
Added Joshua Meier, Absci Senior VP and Chief AI Officer “This is our moonshot effort. People were telling us this was impossible to do,” Meier added.
Meier is the corresponding author on a preprint that detailed Absci’s antibody milestone, “Unlocking de novo antibody design with generative artificial intelligence,” posted January 9 on bioRxiv.
In the preprint, Meier and a team of 34 co-authors wrote that while the primary focus of their research was the in silico design of HCDR3, fully de novo antibody design will require the generation of multiple antibody CDR regions.
“We show initial progress toward this goal with a multi-step generative AI approach for designing all three heavy chain CDRs (HCDR1, HCDR2, HCDR3),” Meier and colleagues reported in the preprint. “Taken together, this work paves the way for rapid progress toward fully de novo antibody design using generative AI, which has the potential to revolutionize the availability of therapeutics for patients.”
Meier and colleagues added that their future work “will expand generative design to enable the de novo design of all CDRs and framework regions, further diversifying possible binding solutions.”
“Developing epitope-specificity across multiple antigens for antibody designs could allow for precise interaction with biologically relevant target regions associated with disease mechanisms of action. In addition to advancements on the generative modeling front, the speed and scale of wet lab validation for AI-generated designs will progressively increase as the time and cost of DNA synthesis continue to decline,” Meier and co-authors predicted. “The controllability of AI-designed antibodies will enable the creation of customized molecules for specific disease targets, leading to safer and more efficacious treatments than would be possible by traditional development approaches.”
Absci said the milestone also marked the first instance of a generative AI engine designing new therapeutic antibodies by designing the heavy chain complementarity determining region 3 (HCDR3) from scratch with generative AI methods using trastuzumab and its target antigen, HER2, as a model system.
“We actually went out after the hardest CDR to work on,” Meier said. He noted that of the six CDRs that serve as the key area for antigen binding and the main area of structural variation in antibodies, CDR3 is the region on the antibody that has the highest sequence diversity in immune repertoires and high density of paratope residues. “If you look across different antibodies, it’s the one that’s most involved in binding, and it’s historically been the one that’s hardest to model with machine learning methods as well.”
440,000 Antibody variants
Absci’s lab designed de novo, then screened approximately 440,000 antibody variants designed for binding to HER2 using its proprietary high-throughput Activity-specific Cell-Enrichment (ACE) assay. From these screens, Absci further characterized 421 binders using surface plasmon resonance (SPR) and estimated the presence of approximately 4,000 binders among its designs.
The company’s investigators found three binders that bound tighter than trastuzumab, the therapeutic antibody marketed as Herceptin® by Roche and its Genentech subsidiary, and also available in biosimilar versions marketed by Amgen, Celltrion, Pfizer, Samsung Bioepis, and Viatris (formerly Mylan).
Absci asserts that its antibody breakthrough shows that generative AI can serve as an alternative to traditional biologic drug discovery by potentially unlocking treatments for traditionally “undruggable” diseases and improving therapeutic possibilities for many others.
Absci is among several companies applying AI to design novel antibodies. Last November, British firm Exscientia—which spent a decade pioneering the use of AI toward designing small molecule drugs—expanded its AI-based platform into novel antibody design as a first step toward designing precision engineered and optimized, fully human biologics. Oxford-based Exscientia says its expanded platform will enable it to develop next-generation therapeutic antibodies through generative AI design—and will enable a near doubling of the universe of potential targets for new treatments.
“Exscientia is pleased to hear of ongoing innovation in AI and efforts to develop new ways of developing antibodies,” Andrew Hopkins, PhD, Exscientia’s founder and CEO, told GEN Edge.
“Our work is focused on antibody by design, not discovery, for specific epitopes beyond what is possible through conventional library screening. Because we don’t have knowledge of the computational methods or visibility into this study’s baseline data, we are unable to comment on it directly, though this work points to a broader ongoing evolution in drug discovery in development.”