AI-guided design, robotic synthesis and automated screening are linked into a closed loop that yields improved JAK-family inhibitors over two cycles
- Combines generative molecular design guided by an affinity predictor, AI retrosynthesis, robotic synthesis and automated screening, with expert oversight for some synthesis and purification steps.
- Applied to four JAK-family targets (JAK1, JAK2, JAK3, TYK2) over two design-make-test-analyse cycles, with a few dozen compounds made or purchased and tested experimentally.
- The first round mainly produced analogs of known inhibitors; the second round gave compounds with improved potency and ligand efficiency across all four targets.
💡 Points Based on the abstract
Round one returning analogs of known inhibitors suggests the generator mostly leaned on JAK chemotypes already in the training data, so improved potency in round two may reflect optimizing a familiar scaffold. Without a matched manual or random-selection arm, the loop's added value is unclear. Ask for selectivity data across the JAK paralogs, since potency alone is easy to gain.
Toward full-cycle automation of closed-loop molecular discovery – a case study on JAK targets ↗