Skip to content
Atlas of Computational Cell Reprogramming

← Methods

L2 · Proxy inverse design D P

DrugReflector

DeMeo B, Nesbitt C, Miller SA, Burkhardt DB, Lipchina I, Fu D, Holderrieth P, Kim D, Kolchenko S, Szalata A, Gupta I, Kerr C, Pfefer T, Rojas-Rodriguez R, Kuppassani S, Kruidenier L, Doshi PB, Zamanighomi M, Collins JJ, Shalek AK, Theis FJ, Cortes M

2025 · Science (New York, N.Y.)

Phenotypic drug screening remains constrained by the vastness of chemical space and the technical challenges of scaling experimental workflows.

Abstract

From the original paper, Science (New York, N.Y.) · PubMed

Phenotypic drug screening remains constrained by the vastness of chemical space and the technical challenges of scaling experimental workflows. To overcome these barriers, computational methods have been developed to prioritize compounds, but they rely on either single-task models lacking generalizability or heuristic-based genomic proxies that resist optimization. We designed an active deep learning framework that leverages omics to enable scalable, optimizable identification of compounds that induce complex phenotypes. Our generalizable algorithm outperformed state-of-the-art models on classical recall, translating to a 13- to 17-fold increase in phenotypic hit rate across two hematological discovery campaigns. Combining this algorithm with a lab-in-the-loop signature refinement step, we achieved an additional twofold increase in hit rate along with molecular insights. In sum, our framework enables efficient phenotypic hit identification campaigns, with broad potential to accelerate drug discovery.

Summary

Editorial summary pending review by the maintainer. The paper's own abstract appears above; the Atlas summary in the maintainer's voice will explain how DrugReflector relates to the cross-modality inverse-design framework of the review.

Why this level

Level 2 because the method instantiates the inverse-design objective through an empirical or learned proxy response map F^proxy(PS,u)\hat F_{\mathrm{proxy}}(P_S, u) rather than a mechanistic intervention-dependent model. Representation family is signature / state-matching. Cited in §3.3 of the review. Editorial rationale pending review by the maintainer.

Classification

Level
L2
Representation
Signature / state-matching
Modalities
D, P
Intervention
Transcription factors
Framework
Machine learning

Software

Reproducibility
4 of 4
FAIR4RS
3 of 5

Last audited 2026-05-24

Citation

DeMeo B et al. (2025). Active learning framework leveraging transcriptomics identifies modulators of disease phenotypes., Science (New York, N.Y.).

DOI: 10.1126/science.adi8577

PMID: 41129612

BibTeX
@article{drugreflector2025,
  title  = {Active learning framework leveraging transcriptomics identifies modulators of disease phenotypes.},
  author = {DeMeo B et al.},
  year   = {2025},
  journal = {Science (New York, N.Y.)},
  pmid = {41129612},
  doi  = {10.1126/science.adi8577}
}