About Valence Labs
Valence Labs is Recursion’s AI research lab, focused on building the computational foundations for a new paradigm in drug discovery.
Our core belief is simple: biology can be modeled, not just measured.
For decades, drug discovery has been constrained by the speed, cost, and variability of real-world experiments, from early screening through late-stage trials. While data generation has scaled, our ability to reason over that data and generalize beyond it has lagged behind. Valence exists to close that gap.
We build machine learning systems that turn biological and chemical data into predictive, mechanistic, and ultimately generative models of living systems. The goal is not just to analyze experiments but to simulate them, shifting the role of the lab from discovery to validation.
What We’re Building
Our work spans three tightly connected layers of the discovery stack:
Biological systems
We are developing large-scale models that simulate how cells respond to perturbations across gene expression, morphology, and beyond. These systems, known as Virtual Cells, aim to predict the effects of interventions before they are tested, enabling virtual experimentation at scale.
Physics-informed chemical systems
We build models that integrate physical priors with machine learning to better predict molecular interactions, especially in regimes where data is sparse or biased. This allows us to generalize to new targets and improve early-stage decision-making in molecular design.
Autonomous discovery systems
We are assembling end-to-end systems that combine generative design, predictive modeling, and simulation into closed-loop discovery engines, expanding the space of ideas explored while increasing the quality of decisions made.
Why It Matters
Drug discovery remains fundamentally limited by iteration cycles of design, test, learn. We believe the next step change comes from compressing this loop to predict before you test, allowing for the exploration of orders of magnitude more hypotheses and higher-confidence decisions, earlier.
By building systems that can simulate biology and chemistry with increasing fidelity, we aim to make discovery faster, more reliable, and more scalable.
How We Work
Valence Labs operates at the intersection of research and deployment.
We publish and contribute to the scientific community, but we also build systems that are integrated into Recursion’s platform and used to drive active discovery programs. Progress is measured not just by machine learning benchmarks but by impact on real drug discovery outcomes.
Our team brings together machine learning researchers, computational biologists, chemists, and engineers working as one integrated group, turning advances in AI into tangible improvements in how drugs are discovered.
Looking Forward
We’re still early but the direction is clear: moving from reactive experimentation to predictive simulation, from fragmented tools to integrated systems, and from data accumulation to understanding.
Valence Labs is building toward a future where we can model biology with enough fidelity to design interventions with intent before they ever enter the lab.
