A recent presentation at the 2026 ASGCT meeting in Boston highlighted Lila Sciences’ autonomous science platform, which has demonstrated significant advancements in mRNA stability and CAR-T therapy development. The platform, which has tested over 950,000 mRNA sequences, achieved deeper and more durable B-cell depletion in non-human primates with an in vivo CAR-T therapy, outperforming a leading benchmark composition.
This breakthrough was accomplished by a team of three Lila scientists in just six months, at a fraction of conventional development costs. The platform’s ability to rapidly design, build, and measure mRNA sequences in living cells, feeding results back into a scientific reasoning model, is changing the landscape of drug development.
Messenger RNA (mRNA) therapies, while promising, face challenges due to the fragility of mRNA molecules, which degrade quickly in the body. This short functional lifespan limits protein production, affecting dosage, cost, and potential side effects. Lila Sciences’ platform addresses this by identifying and testing more durable mRNA sequences. For instance, the platform developed mRNA sequences for erythropoietin (EPO) that produced as much protein on day 15 as benchmark mRNA did on day 2. Similar robust stability was observed for other therapeutically relevant proteins like Factor IX and IL-10.
Ben Kompa, Vice President and Head of AI Lab Innovation at Lila Sciences, stated that the platform is designed to enable scientists to move from a prompt to potentially high-value therapeutic designs. The vast design space for optimizing mRNA’s multiple functional regions makes traditional lab workflows inefficient, often leading to testing only a small number of candidates. Lila’s platform tackles this combinatorial problem by running the full cycle of hypothesis, experiment, and iteration at an unmatched speed and scale.
The AI model is trained on a proprietary dataset of 950,000 RNA sequences that have been physically synthesized and tested in Lila’s AI Science Factories (AISFs). This allows the AI to learn the rules of RNA biology from real experimental outcomes, rather than just published literature. When prompted with a therapeutic objective, the AI identifies relevant biological sequences and designs candidate mRNA molecules, co-optimizing various regions simultaneously. Yue Yang, a molecular biologist at Lila who led the mRNA work, noted the platform’s ability to generate new hypotheses and explore novel sequence spaces rapidly.
The 5’UTR and 3’UTR regions of mRNA are crucial; the 5’UTR controls translation initiation, and the 3’UTR governs stability and degradation rate. The codon region dictates ribosome flow. Getting these components right is essential for sustained protein expression. Lila’s AI model proposes strong candidate sequences from the outset, reducing the need for extensive wet lab runs. The continuous feedback loop from thousands of experiments in the AISFs refines subsequent designs, creating a compounding effect of better designs and faster testing.
The success with EPO was not an isolated incident. Lila’s sequences consistently outperformed published benchmarks from major pharmaceutical companies across five additional protein payloads in follow-on experiments measuring mRNA half-life in human cells, including Factor IX, IL-10, IL-22, eGFP, and firefly luciferase. Kompa emphasized the consistent validation of ultra-stable protein expression.
The application of the platform to CAR-T therapies began in July 2025, following a challenge from CEO Geoff von Maltzahn. Conventional CAR-T therapy, which involves ex vivo engineering of a patient’s immune cells, is expensive, slow, and logistically demanding. In vivo CAR-T, which delivers CAR-encoding mRNA directly into the body, offers a more accessible alternative. Within two weeks, Lila’s scientists designed anti-CD20 CAR mRNAs, formulated them, and began testing. The results in non-human primates showed peripheral B cells depleted to near-zero by day 4, maintained through day 14, a significant improvement over benchmark sequences that rebounded by day 14. This deeper and more sustained depletion could lead to better disease control and reduced treatment needs.
Bob Gantzer, Vice President of Next Generation Platform at Lila, highlighted that the platform is designed to accelerate research to translation. Lila is expanding its platform, with a new 270,000-square-foot AISF in Alewife, Massachusetts, designed for continuous, 24-hour experimental cycles. This expansion aims to explore vast regions of RNA design space, potentially leading to new medicines with improved efficacy and fewer side effects.
