Most leads originate from known compounds or from high-throughput screens, found a recent report in the Journal of Medicinal Chemistry on newly disclosed clinical candidates.

Dean Brown and Jonas Boström, researchers at AstraZeneca, analysed the origins of 66 clinical candidates that were first reported in the Journal of Medicinal Chemistry in 2016–2017. They found that 43% of these molecules were derived from previously known compounds, 29% originated from high-throughput screening of large random compound libraries, 14% came from structure-based drug design strategies, 8% came from directed screens of small and well-characterized compound libraries, 5% came from fragment screens and 1% came from DNA-encoded libraries. Some, but not all, of the drugs that originated from known compounds were fast followers.

Few of these new clinical candidates were derived from phenotypic screens, they report. A 2011 analysis in Nature Reviews Drug Discovery on the origins of 259 FDA-approved drugs found that target-based screening was the most successful lead-generation strategy for follow-on drugs, whereas phenotypic screening was the most successful approach for first-in-class drugs.

Brown and Boström also assessed the physicochemical properties of hits and their respective candidate drugs, and found that molecular weight was typically the only significant change. Lipophilicity, as measured by cLogP, did not change on average. More than 50% of the clinical candidates were structurally very different, and more complex, than their starting points. But in one case — Gilead Sciences' remdesivir (GS-5734), a nucleoside analogue for the treatment of Ebola — the drug candidate was discovered directly in the screen.

“This study provides a retrospective analysis of past trends which have led to successful clinical candidates, and hopefully proves the framework for the exploitation of future opportunities,” the authors write.