Answers are grounded in this paper and cite their sources.
Here, we present a reaction optimisation workflow that integrates data mining to generate unbiased, data-led starting points, high-throughput experimentation (HTE) to narrow the reaction parameter space, and closed-loop flow experiments guided by multi-objective Bayesian optimisation for fine-tuning. Designed for early-stage medicinal chemistry, particularly fragment-based drug discovery (FBDD), the workflow was demonstrated in two direct two-step syntheses of potent fragment-derived protein kinase C inhibitor precursors, involving sequential C–C and C–X bond formations. Quantum-mechanical (QM) reactivity prediction was used to inform route redesign improving synthetic efficiency and safety. HTE identified promising categorical starting points, which were subsequently refined within 20-40 autonomous continuous flow experiments per campaign, across 4 reaction steps. The workflow improved reaction yields in both syntheses while minimising simple E-factor and incorporating selectivity as an additional optimisation objective to facilitate downstream purification. An accelerated variant identified comparably productive conditions within a single working week, reducing substrate consumption by approximately 60%. This work demonstrates how staged, data-informed optimisation can accelerate material delivery in early-stage drug discovery while reducing experimental burden and providing valuable insights for subsequent process optimisation.
No takes yet. Share an insight, caveat, or question.
Soritz et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: