Commentary on modern approaches to phenotypic drug discovery

Another paper from Lilly this year (who have been pushing widely to raise the profile of their open innovation phenotypic and target based screens highlights their recent success in identifying both new chemical hit matter and new molecular targets for an old indication – angiogenesis.

To recap, the usual cries of desperation generated by the well-worn facts and figures about declining pharma productivity in the face of rocketing costs have caused a re-evaluation of the classical phenotypic drug discovery paradigm vs. the single molecular target-based screening vs. the so-called neoclassical modern phenotypic approach.  The main weight behind re-engaging with phenotypic approaches comes from analyses of new small molecule FDA approvals from 1999 to 2008, which in last year’s Swinney review showed that for first in class approvals 37% derived from phenotypic screening and 23% from target-based screening, despite a significant bias across the industry for the latter.  Factors often cited for a reluctance to engage in phenotypic screening include assay formats, statistical reliability, SAR generation, target elucidation etc., although in particular for the last point, regulatory guidelines do clearly not require identification of the molecular target for either initiation of clinical trials or molecule approval.  Interestingly, 29% of the new small molecules approved between 2001 and 2004 did not have a known molecular target (  Furthermore, to contradict the existing dogma that target-based screening delivers a more selective, and hence safer, mechanism, an analysis of over 890 approved drugs indicated that over 788 share at least 1 molecular target with another drug (for data that exists to date), with the average drug hitting 6 known molecular targets.

The details of the specific investigation in this paper focus on a phenotypic screen looking for new anti-angiogenesis agents to improve on the current standard of care (VEGFR / Tyr kinase inhibitors) by running a phenotypic screen for endothelial cord formation – specifically eliminating any known kinase actives.  From a screen of 32000 compounds, the 4% hit rate was bolstered by standard hit expansion techniques to yield a total of 451 confirmed actives (with usual filters etc. during confirmation).  Of these, with known biochemical activity, the majority were kinase inhibitors, then GPCR, NHR and finally PDE ligands.

Over 40% had no known biological activity.  The onward activity then focussed on (1) identifying known molecular targets from data mining of known biological activity – and by doing so highlighted a number of ‘likely suspects’ from angiogenesis and cell cycle pathways, as well as the less expected acetyl CoA carboxylase and functional beta-sec and (2) successfully establishing SAR in the absence of a molecular target.

The structures and data are all well described in the paper, but are only of specific interest for those pursuing this therapeutic area – however, the principles of this phenotypic assay delivering robust, statistically meaningful data to drive a discovery project are well made.