Relation reveals MORGAN, a cellular model, along with a $110 million agreement with GSK.

      The quest to create an AI "foundation model of the cell" faces a significant challenge due to a lack of sufficient, reliable biological data for training. Relation Therapeutics believes that generating this data itself is the solution, and GSK is investing up to $110 million to support this endeavor.

      On Thursday, the London-based startup announced two major developments. The first is MORGAN, their primary cellular foundation model, and the second is an expanded partnership with GSK to produce the data required for it.

      MORGAN, which stands for Multi-Omic Regulatory Genomics using Artificial Neural Networks, is designed to predict how human cells will respond to genetic modifications or medications. If successful, this model can identify disease mechanisms and potential new drug targets virtually, prior to any laboratory experimentation. This concept aligns with the foundation model development for other systems, such as the chest, or the AI-biology initiatives at DeepMind's Isomorphic.

      Relation's focus extends beyond the model itself to include the underlying data. The effectiveness of such models depends heavily on the quality and quantity of data they are trained on, and currently, human biology data is limited. To address this, Relation is establishing automated laboratories that conduct cellular experiments with what it terms “superhuman consistency,” generating “petascale” datasets that illustrate cell responses to various interventions over time.

      GSK's involvement is centered on this data production facility. Through their agreement, Relation could receive up to $110 million in upfront and milestone payments. In exchange, Relation will produce extensive human “perturbation” data, which will be integrated into models like MORGAN. These datasets illustrate how cells react to genetic alterations and pharmacological influences, which GSK aims to utilize in identifying new drug targets.

      David Roblin, the CEO of Relation, stated, “The future of drug discovery depends on our ability to understand how human cells respond to disease and treatment at an unprecedented level of detail.” He refers to MORGAN and the associated laboratory as “the scientific infrastructure” for the next generation of drug discovery.

      Relation is not the sole contender in the pursuit of biologically-based modeling. Competitors, including well-funded entities like Chai Discovery and numerous AI-bio labs, are vying for the same goals. Pharmaceutical companies are rapidly establishing AI research collaborations.

      The stakes are high. Relation informed Endpoints News that their methodology could yield “transformationally different success rates” in drug discovery, though this remains unverified. The majority of drugs still encounter failure during trials, and as of now, no cell foundation model has led to an approved medication. The GSK partnership indicates that major pharmaceutical companies are now willing to invest in data generation, rather than solely focusing on the drugs themselves.