GSK and Relation Therapeutics have expanded their artificial intelligence drug discovery collaboration through an agreement worth up to $110 million.

According to Relation's July 30 announcement, the agreement combines upfront and success-based milestone payments. The companies plan to generate large-scale datasets describing how human cells respond to genetic changes and drug interventions.

Those experimental datasets will be used to train artificial intelligence foundation models intended to improve understanding of disease biology and support the identification of potential therapeutic targets.

Experimental data at the center

Reuters reported that the collaboration will include Relation's MORGAN platform. Rather than relying primarily on biomedical text, the models will be trained using structured measurements of cellular responses to genetic and pharmacological interventions.

Relation describes MORGAN as a cellular foundation model developed using large-scale multi-omic perturbation data. In a separate company announcement, Relation said the platform is intended to help researchers examine disease mechanisms, identify potential targets, and guide early drug discovery.

These remain research objectives. The announcements do not establish that MORGAN has discovered a medicine or improved the probability that a candidate will succeed in clinical development.

Building on an existing partnership

The agreement extends an existing relationship between the companies. In December 2024, Relation and GSK announced collaborations focused on identifying and validating targets for fibrotic diseases and osteoarthritis.

The expanded work places greater emphasis on generating reusable human-cell datasets and training foundation models. The companies have not identified a specific drug candidate arising from the new agreement.

The $110 million figure represents the maximum potential value of the collaboration, including contingent success-based payments. It should not be interpreted as an immediate payment or acquisition price.

The connection to Health AI

The collaboration illustrates how Health AI is being applied upstream in the pharmaceutical pipeline. Here, AI is being used to organize and model experimental biology before a potential medicine reaches preclinical or clinical development.

The approach also reflects a broader shift toward building models and experimental systems together. Laboratory results provide new training data, while computational models can help researchers prioritize the next biological questions to investigate.

Target identification remains an early research step. Any resulting hypothesis would still require laboratory validation, preclinical testing, clinical trials, and regulatory review before contributing to an approved treatment. The announcement is a research collaboration, not evidence of a new medicine or a clinical result.