Nabla Bio Generated Antibody Candidates With JAM-2 Model
The company reported that its AI model successfully designed antibodies against 26 distinct targets.
Updated on Sept. 28, 2026 in Biotech
Live Poll
Do you trust AI-developed drugs to be as safe and effective as traditional medical therapies?
Nabla Bio announced that its JAM-2 AI model has generated antibody candidates against 26 targets, achieving up to an 11 percent success rate for on-cell binding against G-protein-coupled receptors (GPCRs). The development, which remains in the research stage, also produced candidates where over 50 percent met developability standards without additional optimization.
Why it matters
By automating the discovery of antibodies that can activate specific cellular signaling pathways, this AI-driven approach seeks to accelerate the timeline for developing complex biological therapeutics. The ability to hit user-defined epitopes at scale potentially lowers the costs and time required for preclinical drug screening.
The JAM-2 model achieved a 30 to 70 percent precision rate for user-defined epitopes during testing. The model utilizes test-time scaling—a method that iterates through multiple rounds of internal reasoning—to refine its molecular outputs before finalizing candidates.
The players
Nabla Bio
A Massachusetts-based biotech firm focused on using AI models like JAM-2 to generate protein and antibody therapeutics.
Takeda
A global research-driven pharmaceutical company that partnered with Nabla Bio to develop new antibody molecules.
Chai Discovery
An AI-focused startup specializing in biological foundation models that recently secured $400 million in funding.
The details
The JAM-2 model works by iterating through rounds of reasoning to select antibody sequences that satisfy specific binding requirements. This process allows the system to design molecules capable of triggering cellular signaling pathways, a function critical for therapeutic efficacy. Over half of the generated candidates met standard developability criteria, which refers to the biophysical properties needed to successfully manufacture and store a drug, without requiring further laboratory optimization.
Timeline
October 2025: Nabla Bio and Takeda announced a partnership expansion.
Mid-2026: Chai Discovery raised $400 million in funding.
2027-2028: Nabla Bio expects to conduct first-in-human clinical trials.
The Tech Race
Nabla Bio’s progress with JAM-2 places it among a group of emerging biotech firms utilizing generative AI to compress the timelines for antibody discovery. This trajectory closely tracks the rapid capital influx seen in the field, including the $400 million recent funding round secured by Chai Discovery.
The potential success of this technology could eventually lower drug development costs, as successful molecules created through the Takeda partnership may trigger up to $1 billion in payments to Nabla Bio. These developments remain in the research phase and will not impact existing patient treatment workflows for several years.
The takeaway
Nabla Bio has demonstrated that generative AI models can meet complex developability criteria for antibody candidates, setting a new benchmark for automated protein design. Observers should watch for the initiation of human trials in 2027 to see if these computational results translate into clinical success.
What happens next
Nabla Bio is currently working toward initiating first-in-human clinical trials for its AI-designed antibody candidates between 2027 and 2028.
Further reading
For more on how computational methods are changing drug design, see the latest updates in Biotech.
Source note: This article includes information reported by Crypto Briefing.
Live Poll
Do you trust AI-developed drugs to be as safe and effective as traditional medical therapies?
