AI competition shows better data beats larger models in drug metabolism predictions

Inductive Bio emerged as the winner of a recent drug metabolism AI competition, but the final standings revealed a 28-way statistical tie, underscoring that the quality of data can outweigh model size.

The tie suggests that richer, more relevant datasets may be more critical than simply scaling up AI models when predicting how drug candidates interact with metabolic pathways.

For drug developers, the findings are especially relevant to the pregnane X receptor (PXR), a sensor that triggers the production of enzymes responsible for breaking down roughly half of all marketed medicines. Accurate early predictions of PXR activation could help avoid costly late‑stage failures.

The results add to a growing body of evidence that AI tools must be paired with high‑quality data to deliver practical benefits in pharmaceutical research.