GSK and Relation Therapeutics expanded their AI pact, with GSK set to pay Relation up to $110 million to scale the discovery of therapeutic targets using multi-omic perturbation datasets. The collaboration focuses on generating high-quality, time-resolved datasets capturing how human cells respond to genetic and pharmacological interventions. Relation will lean on automation to generate large, consistent multi-omic readouts for training foundation models intended to improve target discovery confidence and accelerate biological validation. The company positioned the work as a “data problem” as much as a modeling effort. The deal also arrives alongside Relation’s launch of MORGAN, its foundation model of cellular perturbation response across disease contexts. That model is framed as general-purpose, spanning cell types and disease areas rather than being tuned to a single indication. For biotech platforms, the expanded pact underlines the increasing use of foundation-model training pipelines built from proprietary, experimentally generated multi-omic datasets—where output quality and reproducibility are becoming the differentiators.