Recursion Pharmaceuticals executives provided an overview of the company’s artificial intelligence-driven drug discovery platform during a Morgan Stanley webcast, explaining how the technology is being used across target identification, molecule design and clinical development.
Chief Executive Najat Khan and Chief Financial Officer Ben Taylor described a workflow that couples machine learning with experimental biology to chart the effects of genetic perturbations across biological systems. Management cited a mapping project that simultaneously uncovered 25 new targets, a result the company said condensed work that would traditionally unfold over years of sequential research.
On the chemistry side, Recursion detailed progress in molecule design. The company reported advancing one compound into clinical trials after designing about 330 molecules over a 17-month period. By contrast, management said typical industry practice involves designing roughly 2,500 to 5,000 compounds over four to five years. The firm leverages generative artificial intelligence to optimize candidate molecules for potency, selectivity, safety and synthetic feasibility prior to laboratory synthesis.
Executives also discussed operational uses of AI in clinical development. Management said that AI-driven methods are already speeding patient enrollment by 30% to 50% through improvements in recruitment and site selection, and that the company applies these tools to protocol design and patient stratification as well.
Recursion has assembled a substantial experimental data asset, reporting 40 petabytes of proprietary data. That repository is supplemented by an additional 25 petabytes obtained through partnerships, including one with Roche’s Genentech unit, the company said.
Regarding development costs, management observed that animal testing and related chemistry, manufacturing and controls work can represent roughly half the cost from initial concept to a development candidate. They said Recursion expects millions in savings as its predictive models mature and reduce the need for some traditional experimental steps.
The executives also highlighted the opportunity space: they said approved and late-stage drugs currently address only 10% of the genome, and that clinical development consumes approximately 70% of the capital required for successful drug development.
Summary
Recursion presented details on how machine learning and experimental biology are integrated across its discovery and development process, citing specific examples of compressed timelines in target mapping and molecule design, quantified data assets, partnership-augmented datasets and expectations for cost reduction as predictive models improve.
Key points
- Recursion combined machine learning and experimental biology to map genetic perturbations, identifying 25 new targets in a single effort - impacts drug discovery and biotech R&D sectors.
- The company advanced a compound to clinical trials after designing ~330 molecules over 17 months versus typical industry ranges of 2,500-5,000 over 4-5 years - relevant to pharmaceutical development timelines and efficiencies.
- AI tools are reported to accelerate patient enrollment by 30%-50% and are used for protocol design and patient stratification - affecting clinical operations and trial economics.
Risks and uncertainties
- Only 10% of the genome is addressed by approved and late-stage drugs, indicating scientific and target discovery risk across the biopharma sector.
- Expected cost savings rely on the continued development and predictive accuracy of models; the scale of savings is described as "millions" but depends on model performance.
- Clinical development accounts for roughly 70% of the capital required for successful drug development, creating financial exposure for programs as they move through costly clinical stages.