ALINA: Active Learning Inference for Nucleic Acid Delivery

Objectif

RNA therapeutics hold enormous potential, yet progress is hindered by slow, costly, and empirical formulation workflows that translate poorly from in vitro to in vivo and complicate freedom-to-operate (FTO) decisions. Current strategies rely on trial-and-error, one-factor-at-a-time approaches that deliver incremental gains and rarely unlock extrahepatic delivery. ALINA will establish the technical and commercial proof-of-concept for an integrated machine learning (ML) and molecular dynamics (MD) platform that accelerates the design and optimization of mRNA-loaded lipid nanoparticle (LNP) formulations. The platform combines curated multi-modal datasets, physics-informed MD descriptors, active-learning experiment selection, and higher-translational biological models to identify globally optimal, FTO-conscious LNP compositions with significantly fewer experiments. The project will deliver (i) an industry-compatible MVP of the ALINA platform, (ii) quantitative benchmarks demonstrating faster convergence and improved predictive accuracy versus current methods, and (iii) an IP strategy covering newly identified lipid components and a commercialization plan combining platform-access services and licensing. By improving development speed, predictive power, resource efficiency, and route-to-impact beyond the liver, ALINA addresses a major bottleneck in RNA therapeutics development and supports the broader adoption of data-driven formulation across biotech and pharma that could benefit at least 2-4 million patients annually.