Core Technology
DTIGN — drug-target graphs, at industrial scale.
A graph-based deep learning stack that turns a 2B+ compound library into actionable, ranked candidates for our preclinical programs.
Key metrics
- Compound library
- 2B+
- Molecules generated
- 10M+
- Hit-rate uplift
- 64x
- Predictive accuracy
- 44.83%
Virtually screenable molecules accessible through our AI pipeline.
Novel candidates produced by our generative discovery models.
Measured improvement over conventional high-throughput screening.
Top-line model performance reported in our IEEE DTIGN paper.
Generative chemistry
10M+ novel candidates generated, filtered through pharma-grade ADME/Tox screens before in vitro validation.
Drug-target graph network (DTIGN)
Our IEEE-published model encodes drug-target topology to predict binding, achieving 44.83% accuracy on benchmark assays.
Closed-loop validation
Wet-lab assays feed back into the model weekly, compounding hit-rate improvements (currently 64x over HTS).
Publications
DTIGN: Drug-Target Interaction Graph Network
Nanyang Biologics Research · IEEE, 2024
Research Briefs
Subscribe to NYB Research Briefs
Monthly digest of our AI drug discovery progress, publications, and longevity science.