GSK and Relation Therapeutics expand AI drug discovery partnership

Relation’s approach centers on its Lab-in-the-Loop platform, which merges computational analysis with wet-lab experiments. By generating large-scale, specialized datasets—such as the proprietary Osteomics bone atlas—the company attempts to bypass the limitations of public repositories, which often suffer from technical noise, batch effects, and redundant data that can skew model training.

Recent industry research suggests that massive data volume is no longer the sole driver of performance. Studies in Nature Methods and Genome Biology indicate that single-cell foundation models frequently hit performance plateaus, suggesting that model architecture and data quality are more critical than raw size. This realization is shifting the strategy of major pharmaceutical players, who are increasingly favoring partnerships that provide access to high-quality, disease-specific data over simple licensing of general datasets. Alongside the GSK deal, recent industry moves—such as AstraZeneca’s $200 million agreement with Tempus—underscore a broader trend where companies prioritize curated, proprietary datasets to overcome the primary bottleneck in AI-assisted drug development.