Reflector Bio emerges from stealth. Read the announcement

In silico phenotypic screening

Prevailing methods are structure-based (AlphaFold, Boltz), but they don't account for cellular context, off-target effects, or mixed effects in vivo. When those surface later, it's back to the drawing board.

But what if you could see the downstream effects of your molecules, earlier?

We start with in silico secondary and cell-based assays: predicting efficacy and giving you the full picture for go/no-go decisions.

Drug discovery began with an effect, not a target

Phenotypic screening, testing a compound for a desired biological outcome directly, is how most drugs were found until about twenty years ago, when genome-wide association studies (GWAS) made it possible to identify a validated target before ever touching a compound. The field shifted toward target-based discovery, and mostly stayed there.

01

It had a resurgence, then stalled again

Phenotypic approaches saw a comeback in the 2010s. Approval and pharma M&A frameworks built around a known target and mechanism made it hard to advance a phenotypically discovered compound, and running the screen at scale stayed too resource-intensive for most teams.3

02

It still finds drugs target-based methods miss

In an analysis of 75 first-in-class drugs approved between 1999 and 2008, phenotypic screening produced 28 of them, more than target-based approaches, which produced 17.1 It captures polypharmacology and surfaces hits serendipitously, in ways structure-based docking cannot, since docking requires committing to a target and a binding pocket before you even start.2

03

Our goal is to make it viable everywhere

We built Reflector to remove both blockers, so any biotech or pharmaceutical can work in the phenotypic paradigm and translate biological insights into therapeutic innovation at the scale modern discovery requires. We work with teams already screening phenotypically, and with target-based teams running into efficacy or safety problems downstream: a validated target does not by itself guarantee success in human translation.

References
  1. Swinney DC, Anthony J. How were new medicines discovered? Nat Rev Drug Discov. 2011;10(7):507-519. nature.com/articles/nrd3480
  2. Moffat JG, Vincent F, Lee JA, Eder J, Prunotto M. Opportunities and challenges in phenotypic drug discovery: an industry perspective. Nat Rev Drug Discov. 2017;16(8):531-543. nature.com/articles/nrd.2017.111
  3. Scannell JW, Blanckley A, Boldon H, Warrington B. Diagnosing the decline in pharmaceutical R&D efficiency. Nat Rev Drug Discov. 2012;11(3):191-200. nature.com/articles/nrd3681

Where phenotypic prediction fits

Our multimodal foundation model is trained on functional screening data across cell types, assay modalities, and small molecule classes. It is complementary to structural and target-based tools: a distinct axis of information that gives a deeper biological understanding of how compounds act in cells.

Capability
Target-based / Mechanistic / FEP+
Virtual cell
✓ Reflector
Virtual screening of billions of small molecules
No
Yes
Yes
Predicts functional cell response
No
Partial
Yes
Captures polypharmacology
No
Partial
Yes
Ingests multimodal functional assay data
No
Partial
Yes
Works without a known target
No
No
Yes
Predicts binding affinity
Yes
No
No
10M+
Functional experiments
Largest phenotypic model to date
60×
Reduction in screening volume
vs. traditional HTS
80%
Hit discovery rate
Experimentally validated
<4 wks
Data to prioritized candidates
vs. 12+ month cycle

Questions and evidence

FAQs

Frequently asked questions

One-sentence answers on data requirements, fit, and how we compare to structure-based tools.

Read the FAQs →
Case studies

Where phenotypic prediction changed the outcome

Programs where our model triaged hits, flagged liabilities, or reopened archived data.

See case studies →