Generate:Biomedicines is advancing an AI-designed asthma drug into two global Phase 3 trials, marking one of the first large-scale clinical tests of generative biology. The Somerville, Massachusetts company said it will begin the SOLAIRIA-1 and SOLAIRIA-2 studies of GB-0895, a long-acting monoclonal antibody targeting thymic stromal lymphopoietin, in roughly 1,600 patients with severe asthma who are not adequately controlled on standard treatments. It is the company’s most advanced program and the first AI-generated biologic to reach this stage with a global, year-long dataset.
Generate, founded by Flagship Pioneering in 2018, describes itself as operating at the intersection of machine learning, biological engineering, and medicine. The company says its platform uses models trained on millions of proteins to learn the rules that encode biological function, then generates new proteins with therapeutic properties. It refers to this approach as “Generative Biology.” GB-0895 is among the first molecules built fully within this framework to reach late-stage development.
The company has raised nearly $700 million in equity financing across its 2021 Series B and 2023 Series C rounds. It entered a multi-target collaboration with Novartis in 2024 that included $65 million upfront and the potential for more than $1 billion in milestones and royalties. With GB-0895 now advancing to Phase 3, the company is moving from platform claims to validation in a clinically regulated environment.
A Late-Stage Test for AI Drug Development
AI has been used in the pharmaceutical industry for more than a decade, but most activity has remained early-stage: target identification, molecular screening, or small Phase 1 studies. Published reviews point to advantages including faster molecular design cycles, improved prediction of binding affinity, and early modeling of toxicity and pharmacokinetics. Yet few AI-generated biologics have entered large, multi-country trials where efficacy must be demonstrated over a year or longer.
The Phase 3 studies for GB-0895 will measure the reduction in annualized asthma exacerbation rate over 52 weeks. Secondary endpoints include lung function, symptom control, and quality-of-life metrics. According to public disclosures, the antibody is engineered for subcutaneous dosing every six months, a longer interval than current biologics targeting the same pathway.
Earlier data came from a Phase 1 study in 96 patients with mild to moderate asthma. Results showed dose-proportional pharmacokinetics and an estimated half-life of about 89 days. Biomarker reductions consistent with TSLP blockade were sustained for up to six months, supporting the dosing approach being tested in Phase 3.
The company cites the speed of development as evidence for the platform’s impact. Work on the molecule began four years ago, and the company chose to run its Phase 1 study directly in asthma patients rather than healthy volunteers, allowing it to bypass a Phase 2 trial. Chief Medical Officer Laurie Lee said Phase 1 showed the molecule performed according to its design parameters.
There is skepticism in the sector about AI-driven therapeutic claims. In February 2025, STAT reviewed public claims by several AI-native biotech firms, including Absci and Generate, and questioned whether the design contributions of AI systems were being overstated relative to traditional methods and whether accelerated early timelines have any bearing on outcomes in later stages of development. GB-0895’s Phase 3 program will provide the first dataset from an AI-designed antibody at a scale large enough to address those concerns directly.
For Generate, the Trial Is a Proof Point and a Risk
Generate’s platform is structured around a continuous loop. Machine-learning models create protein sequences optimized for specific molecular properties. These proteins are then built and experimentally tested at scale, and the data are fed back into the model to refine future design cycles. The company says it has generated, built, and characterized more than 42,000 proteins, including antibodies, enzymes, cytokines, and “stealth proteins” engineered for reduced immunogenicity.
The company’s broader pipeline includes GB-0669, an AI-generated antibody targeting the SARS-CoV-2 S2 domain. Generate reported in 2024 that the molecule moved from computational design to clinical testing in 17 months and has since completed first-in-human studies. Other programs in immunology and oncology are in preclinical or IND-enabling development.
The outcome of the GB-0895 Phase 3 trials will determine whether Generate’s claims about faster development and higher predictability can be demonstrated in a real-world, year-long clinical setting. Success would position generative biology as a credible model for developing large-market biologics. Failure would clarify the limits of AI-designed proteins and would likely influence how investors and partners evaluate similar platforms.
Generate is treating the trial as a central test of its approach. Chief Executive Mike Nally told the Business Journal he sees the program as “an important moment for the broader AI/ML field” and said AI has improved time and cost to reach pivotal data. The company will have to demonstrate that improvements in design and development timelines translate into clinical performance at scale.
GB-0895 is expected to release Phase 3 readouts after the completion of the year-long trials. Until then, the molecule represents a significant test of whether generative AI can produce biologics that meet the standards of global clinical practice.
