Deep Origin awarded $31.7M ARPA-H contract to replace animal testing with in-silico models. Learn more

Validated assets

We do not just run the models.
We run discovery programs.

These are not drugs we are developing in-house. They are vetted assets we have advanced through our discovery platform across oncology, cardiovascular, inflammation, and metabolic disease — available for out-licensing and co-development with partners ready to take them forward.

Available assets GPR75 spotlight CD73 case study
Available assets

Six assets across four therapeutic areas — available for partnership.

Each is a vetted target or hit series advanced through our discovery platform. Partners take them from here. Spotlights below carry the detailed content.

Engagement model

Out-licensing or co-development — not internal clinical development. We continue work on assets where partners want us to (or add new programs at partner request); the goal is dealable chemistry, not a drug-development pipeline of our own.

Target / Asset Therapeutic area Discovery stage Partnering status
GPR75 Lead asset Obesity / Metabolic Virtual Screening Lead asset, available for out-licensing
PCSK9 Cardiovascular Virtual Screening Actively screening — available for discussion
CD73 / 5NTE Oncology Hit-to-Lead Case study ready — available
IRAK4 Inflammation Hit Expansion Available for discussion
IL17A Inflammation Hit Expansion Available for discussion
Factor XI Cardiovascular Hit Expansion Available for discussion
Lead asset spotlight

GPR75 — an orphan GPCR with human genetic validation for obesity.

The target

GPR75 is an orphan GPCR validated by human genetics as a negative regulator of body weight. Loss-of-function variants associate with significantly lower BMI in large population studies, with no adverse effects in carriers — among the strongest human-genetic obesity signals published in the last decade.

No approved drugs. Open IP. Virtual screening underway.

What this opens up

  • Human-genetic validation — population-scale loss-of-function evidence rather than animal or phenotypic surrogates.
  • Open IP landscape — no approved drugs at this target; first-in-class window without competing programs to navigate around.
  • A new mechanism beyond GLP-1 — complementary to or differentiated from incretin-based programs depending on combination strategy.
  • Tractable on a novel target — limited structural precedent for an orphan receptor would normally constrain foundation-model dockers; our platform retains predictivity at 0–20% Tanimoto similarity (see Why Deep Origin).
Request the GPR75 data package
Case study

CD73 — a 30× improvement over the industry hit-rate benchmark.

CD73 (5′-nucleotidase NT5E) — cell-surface enzyme converting AMP to adenosine; modulates immune suppression and tumor metabolism. Widely expressed in cancers.

Challenge

Industry screens against CD73 yield hit rates of 1% or less1,2 from hundreds to thousands of compounds synthesized. High-throughput screening is the standard approach — expensive, slow, and generates a lot of chemistry that goes nowhere.

What we did

Applied physics-based docking and multi-parameter computational filtering to a virtual compound library before any synthesis was commissioned.

Result
  • 48 of 159 compounds (30%) with binding affinity below 80 µM.
  • 9 compounds below 10 µM.
  • Dose-response — correlated single-dose inhibition with pIC₅₀ from full curves at both 10 and 100 µM.
  • Chemotype diversity — hits cluster at 0.2 Tanimoto similarity, well outside the training distribution.
  • Drug-like properties — SlogP, AMW, TPSA, and Lipinski HBA / HBD distributions all within drug-space.
30× improvement in hit rate vs. the industry benchmark — translating directly into synthesis cost savings, faster lead identification, and a smaller, better-characterized compound set entering lead optimization.
Tanimoto similarity — identified hits cluster at ≈0.2, outside the training distribution. Compounds advanced are chemically novel.
Ranking of foundational models across 22 TDC tasks — Deep Origin Togo leads on early-ADME tasks; CYP-mediated inhibition remains a tradeoff. Request access for more detailed information →
  1. The Atomwise AIMS Program. AI is a viable alternative to high throughput screening: a 318-target study. Sci Rep 14, 7526 (2024). doi:10.1038/s41598-024-54655-z
  2. Kumar M, Lowery R, Kumar V. High-Throughput Screening Assays for Cancer Immunotherapy Targets: Ectonucleotidases CD39 and CD73. SLAS Discov. 2020 Mar;25(3):320-326. doi:10.1177/2472555219893632.

Open to a deeper conversation?

Detailed data sets, IP positions, and partnership structures are shared under CDA on request. Initial calls are with our BD lead.

Request a CDA & data package Explore the science