ML-Driven Design of Lipid Nanoparticles for In-Vivo CAR-T Therapy
Nir Suissa, Eilam Yeini, Hiba Abu-Hariri, Igor Nudelman, Guy D. Rosin, Lee Goldfryd, Maya Himan, Naama Rosmarin, Dalia Moati, Nir Shahar, Inbal Lupovitz, Anat Shermer, Gilad Gibor, Ronnie Shapira-Frommer, Daniel Stuczynski, Amit Benazraf, Yochai Wolf, Gal Cafri, Avi Schroeder
T-cell-based immunotherapies, such as autologous chimeric antigen receptor T-cell therapies (CAR-T), face significant challenges, including complex supply chain and manufacturing processes, high cost of goods, and limited accessibility due to the need for specialized clinical administration sites and extensive lymphodepletion pre-treatment. In vivo delivery of mRNA offers the potential to address these key limitations and expand access to these live-saving therapies. However in-vivo delivery of mRNA to T-cells remains a significant hurdle.
To address this challenge, Mana.bio has developed a proprietary suite of machine learning (ML) models to predict determinants of tropism, safety, and physicochemical stability attributed to the LNP composition and formulation. Through successive campaigns of LNP optimization based on in-vivo biodistribution of fluorescent reporter and CD19 CAR function, Mana.bio designed passive-targeted (i.e. no targeting ligands) lipid nanoparticles (LNPs) that transfect approximately 15% of T-cells in non-human primates and that demonstrate near complete tumor killing in mice in under 9-months.
Collectively, this work demonstrates the potential of artificial intelligence to accelerate the pre-clinical development process.