Petascale Perturbations: GSK Bets $110M on Relation's MORGAN Foundation Model
Relation Therapeutics announced a $110M GSK research pact and unveiled MORGAN, a cellular perturbation foundation model, on July 30, 2026 — betting that petascale human biology data can sharpen AI drug discovery.
Drug discovery partnerships used to begin with a target and a spreadsheet. Relation Therapeutics' $110 million research pact with GSK, announced July 30, 2026, starts earlier — with petascale human cellular perturbation data designed to train foundation models that understand how cells respond before chemists pick a molecule.
On the same day, Relation unveiled MORGAN (Multi-Omic Regulatory Genomics using Artificial Neural Networks), its foundation model of cellular perturbation response across disease contexts. The GSK deal funds generation of time-resolved, multi-omic readouts at a scale CEO David Roblin says conventional labs cannot match.
Data before molecules
Roblin framed the collaboration as a data problem as much as a modeling one. "Deepening our understanding of the underlying biology of disease starts with richer data," he told BioXconomy, "to be used in models that can give us greater confidence in the discovery of therapeutic targets."
Relation's automated laboratories generate perturbation datasets from physiologically relevant human disease systems. MORGAN is built to be general-purpose across cell types and indications — each tissue-specific model learns how human cells respond to genetic and pharmacological interventions.
What differentiates MORGAN, according to Relation, is owning both the model and the scientific infrastructure to train it. The company claims "superhuman consistency" in high-throughput cellular experiments — a quality claim that will face scrutiny as GSK integrates the data into its pipeline.
A virtuous cycle, not a one-off
The GSK collaboration feeds Relation's internal programs in fibrotic disease, osteoarthritis, and osteoporosis while strengthening MORGAN. Roblin described a "virtuous cycle where the platform strengthens the pipeline, and the pipeline helps make the platform even more predictive."
GSK joins a partnership roster that already includes Novartis and Deerfield Management, each occupying a distinct role. Novartis applies Relation's platform to target discovery in shared areas of interest; Deerfield helps translate discoveries into new companies and programs.
Why GSK, why now
Roblin cited complementary strengths: Relation contributes automated experimentation and cellular foundation models; GSK brings disease expertise and clinical development experience. For GSK, the bet is that AI-native perturbation data can sharpen target confidence earlier than traditional screening cascades.
The deal does not replace GSK's internal AI investments — it adds an external data engine tuned specifically for perturbation biology. Whether MORGAN predictions convert to clinical candidates faster than incumbent workflows will be the metric investors watch.
Relation's July 30 one-two punch — MORGAN launch plus GSK funding — marks another step in pharma's shift from AI as a modeling accessory to AI as an experimental substrate. The receipts are in the data generation capacity, not the press release adjectives.
Sources
- BioXconomy — Relation Therapeutics secures $110m research pact with GSK (July 30, 2026)
- Relation Therapeutics — MORGAN foundation model announcement via BioXconomy coverage (July 30, 2026)