In this post, I reflect on the discovery of clinical candidates, inspired by Dean G. Brown’s insightful 2023 paper, comparing his findings with my own experience.
Brown analyzed 156 successful hit-to-clinical candidate pairs published in the Journal of Medicinal Chemistry between 2018 and 2021. He found that most clinical candidates (59%) stem from known starting points. The rest originate from various screening methods: 21% from random screens (e.g., HTS), 11% from directed screens, 7% from fragment screens, and ≤1% from virtual and DEL screens (a).
Interestingly, structure-based drug design (SBDD) isn’t explicitly listed among the screening methods. Brown notes that it’s challenging to determine if SBDD was the lead finding strategy or simply an enabling technology for advancing initial hits. Nonetheless, 65% of the campaigns utilized SBDD to advance hits (b).
These findings resonate with my experience. In biotech, our programs predominantly started from known entities. At Roche, a significant portion of leads originated from HTS and DEL screening. This bias towards known entities likely stems from reduced risk and accelerated development due to prior clinical knowledge. Speed is even more critical in biotech due to the risk of running out of funding.
Regarding SBDD, all my biotech projects were supported by SBDD, whereas at Roche, I also worked on hits from phenotypic screens where the targets were unknown. This raises the question: Is biotech more focused on known entities and more risk-averse than pharma? Is this my personal experience, or a broader industry trend?
Looking ahead, an intriguing question emerges: Will virtual screening yield more hits with the advent of AlphaFold and new AI models? Time will tell.
Reference
Brown DG. An Analysis of Successful Hit-to-Clinical Candidate Pairs. J Med Chem. 2023 Jun 8;66(11):7101-7139. doi: 10.1021/acs.jmedchem.3c00521. Epub 2023 May 24.
