Our approach
AI-assisted drug design, high-throughput screening and structural biology, combined across disciplines to build a new paradigm for drug discovery.
Three platforms
Three platforms that interlock, forming one chain from design to validation.
AI-assisted drug design
Inspired by the structure-based algorithms of conventional computer-aided drug design, we introduced the concept of structure-based AI drug design.
- A suite of algorithms for the design, optimisation and rapid synthesis of small molecules and for drug-likeness assessment, markedly improving the precision and efficiency of discovery.
- Models are retrained on experimental big data to keep improving their accuracy.
High-throughput drug screening
Drawing on DNA-encoded library (DEL) screening and other advanced technologies, we designed our own high-throughput DEL screening protocols.
- Using our in-house screening platform, we selected major cancer-related signalling pathways and core enzymes, together with key enzymes of viral replication, as targets for multiple programmes.
- This work demonstrated that combining AI with DEL screening is a reliable strategy for drug discovery.
Pathway erosion (ECLIPSE)
Pathway erosion (ECLIPSE) is a new drug design strategy that addresses resistance mutations in anticancer molecules.
- An ECLIPSE molecule is bifunctional: a linker connects a target-protein ligand with a functional-enzyme ligand.
A dual engine:
multiplication, not addition
A DNA-encoded library (DEL) holds up to hundreds of billions of compounds, each carrying a unique DNA barcode. The target protein is added to the pool; molecules that bind are pulled out, and their barcodes reveal which compounds were hits.
A single selection can return thousands of hits, and synthesising and testing them one by one is impractical. This is where AI comes in: the model removes false positives and pinpoints the compound that can become a drug, while wet-lab results are fed back to the model so data and model improve together. We describe this as DEL × AI, rather than DEL + AI.
For Mprosevir® (WPV01) we screened a DEL library of 49 billion compounds. AI shrank more than a hundred DEL hits to 9 candidates in a matter of days; 6 of them showed superior activity, and candidate WLU6937 was selected within six months of cloning the 3CLpro protein. The advanced AI architectures were combined with structural biology, medicinal chemistry and biochemistry into our WAND (Westlake AI for Novel Drug) platform, which is continuously retrained on wet-lab data.
From 49 billion
to 1
Figures are taken from the published research narrative “From 49,000,000,000 to 1” and are all on the public record.
- 49 billion
- compounds in the DEL library screened for Mprosevir®
- 9
- candidates shortlisted by AI from more than a hundred DEL hits; 6 showed superior activity
- 3.5 years
- from first discovery to completed clinical trials
- 1
- the first approved small-molecule drug worldwide to originate from DEL technology