Cellarity reported a new framework in Science describing how integrated transcriptomics and AI models can identify compounds that correct pathological cell states. The paper lays out a systematic approach to read cellular phenotypes via high‑dimensional gene expression datasets and employ machine learning to nominate perturbations predicted to shift cells toward healthy states. Authors detail model training on large transcriptomic compendia, validation across disease models and early experimental confirmations in cellular systems. Cellarity’s framework, supplemented by a company release, promotes a cell‑state centric drug discovery paradigm that emphasizes transcriptomic reversal rather than single‑target modulation — a conceptual shift with implications for target selection, phenotypic screening and biomarker development. Publication in Science elevates the methodology among academic and industry audiences and may accelerate collaborations between AI modelers and wet‑lab teams. The work illustrates how multi‑omics integration and generative models are influencing early‑stage drug discovery and could change prioritization for compound optimization and preclinical pipelines.