Cellarity published a Science manuscript outlining a framework that integrates multi‑omic transcriptomic datasets with AI models to discover cell‑state‑correcting therapeutics. The approach trains models to map disease‑associated cell states and predict compounds capable of reverting pathologic states toward healthy profiles. The company argues the method accelerates target identification and compound selection by focusing on cellular phenotypes rather than single molecular targets. The work includes experimental validation and a computational pipeline, and it was accompanied by a Cellarity press release describing the broader platform. The publication reinforces the trend of coupling high‑dimensional biological data with generative and predictive AI to expand drug‑discovery paradigms beyond classic target‑centric screens.