Predictive Validity in Drug Discovery: What It Is, Why It Matters and How to Improve It
| Author | Jack W. Scannell, James Bosley, John A. Hickman, Gerard R. Dawson, Hubert Truebel, Guilherme S. Ferreira, Duncan Richards and J. Mark Treherne |
|---|---|
| Published | October 4, 2022 |
| Format | Research paper |
| Original | Read the original |
TL;DR — A Nature Reviews Drug Discovery Perspective arguing that drug R&D productivity depends heavily on how well screens and disease models predict clinical utility in humans.
Published in Nature Reviews Drug Discovery in October 2022 (vol. 21), with Jack Scannell as lead author. The paper frames drug discovery as a sequence of decisions and uses decision theory to define “predictive validity” as the correlation between a decision tool’s output (a screen, an animal model, an assay) and clinical utility across therapeutic candidates.
Its central claim is that the detectability of good candidates is extremely sensitive to predictive validity, because promising candidates are rare among the very large number tested — the paper’s image is small oases of safety and efficacy in large chemical and biological deserts. A modest gain in the correlation of a model with human outcomes can matter more than large increases in the number of candidates screened. The authors argue predictive validity is under-managed in R&D because it is hard to measure until projects later succeed or fail, and they discuss ways to evaluate and improve it.
The argument extends Scannell’s earlier work on declining R&D efficiency (the “Eroom’s Law” pattern).
What it is: a peer-reviewed review article in a leading pharma-industry journal. The public link is the author-deposited copy in Oxford’s institutional repository (ORA).
Where this came from
- A copy saved in October 2026.
- Oxford University Research Archive (author-deposited copy): https://ora.ox.ac.uk/objects/uuid:a03f3f9a-c8d0-47b0-9ced-7473fdcf6ca8/files/r8336h262x
- Journal page: https://www.nature.com/articles/s41573-022-00552-x
17 pages. A copy is archived locally against link rot; the header links the original source.