GSK, Relation Therapeutics sign $110 million AI drug discovery deal

GSK is paying Relation Therapeutics as much as $110 million to generate human cellular datasets and train AI foundation models, deepening the drugmaker's push into machine-learning-driven target discovery.

"Deepening our understanding of the underlying biology of disease starts with richer data, to be used in models that can give us greater confidence in the discovery of therapeutic targets that can ultimately yield medicines," David Roblin, chief executive officer of Relation, said.

Under the agreement, Relation will produce large-scale perturbation datasets capturing how human cells respond to genetic and pharmacological interventions. The company will use advanced cellular disease models and automated systems to generate time-resolved data with multi-omics readouts — measuring changes across genomics, transcriptomics and proteomics simultaneously. The resulting data will train Relation's MORGAN foundation model, which stands for Multi-Omic Regulatory Genomics using Artificial Neural Networks, to identify biological pathways linked to disease. The collaboration builds on an earlier partnership between the companies focused on fibrotic diseases and osteoarthritis.

For GSK, the deal adds computational biology capacity as it seeks to strengthen a drug pipeline that has faced patent cliffs on key products. The British pharma giant has been increasing R&D investment while expanding AI use to improve the efficiency and success rate of drug discovery, an industry-wide push that has seen similar partnerships between Big Pharma and AI-native biotechs.

Relation's approach combines wet-lab experimentation with computational modeling. The company's integrated automation platform enables the generation of perturbation data at a scale and consistency that traditional methods cannot match, according to the firm. These datasets capture how human cells change over time in response to specific interventions, providing a dynamic picture of disease biology rather than a static snapshot. This temporal dimension is critical because disease mechanisms often involve complex feedback loops and compensatory pathways that single-time-point measurements miss.

The MORGAN model sits at the center of this strategy. Trained on multi-omic data, the foundation model aims to predict cellular responses to genetic and drug perturbations across disease contexts. This type of predictive modeling could reduce the time and cost of target validation, a historically high-failure-rate step in drug development where more than 80% of preclinical candidates fail to reach the clinic. By training on perturbation data from physiologically relevant human disease systems, MORGAN may identify targets with a higher probability of translating into effective medicines.

The deal structure includes upfront payments and success-based milestones, a common framework in biotech-pharma collaborations that aligns incentives around data generation and target validation. Relation will retain ownership of its platform while granting GSK access to the datasets and models generated through the collaboration. The companies did not disclose the specific disease areas covered under the new agreement beyond noting it builds on their existing fibrosis and osteoarthritis work. Relation is also advancing internal programs in immunology, metabolic and bone disease, according to the company.

Relation Therapeutics, a privately held company, does not disclose financials, but the $110 million deal structure offers a reference point for valuing AI-driven drug discovery platforms. GSK shares trade on the London Stock Exchange, and the company has described AI partnerships as central to its R&D strategy. The deal also confirms Relation's biology-first approach, which the company says differentiates it from competitors that rely primarily on computational screening without experimental validation.

This article is for informational purposes only and does not constitute investment advice.