Our first peer-reviewed paper is out in JCIM (special issue). Using BIOPTIC B1, we searched 40B molecules and in weeks—not years discovered novel LRRK2 binders for Parkinson’s, including sub-micromolar hits and 14 actives overall.
Peer-reviewed milestone: BIOPTIC B1 discovers novel LRRK2 inhibitors
Summary. Our ultra-high-throughput, ligand-based virtual screening system BIOPTIC B1 searched 40B make-on-demand molecules and delivered novel LRRK2 binders for Parkinson’s in weeks—not years. The work is peer-reviewed and published in JCIM (special issue).
BIOPTIC B1 turns known ligands into potency-aware embeddings and performs brute-force similarity search over billions of compounds, enabling fast scaffold hopping and patent-friendly novelty without 3D structures.
At a glance
- Scale: 40B Enamine REAL Space compounds screened
- Build speed: 134 predicted leads synthesized in 11 weeks (93% success)
- Results: 14 binders confirmed; best Kd = 110 nM (sub-µM)
- Expansion: 10/47 analogs hit (21% hit rate)
- Novelty: ≤ 0.4 Tanimoto vs any active in BindingDB
- Throughput: CPU-only retrieval over 40B in 2:15 / query; est. compute ~$5 per screen
What makes B1 different
- Search-engine DNA. SMILES-based transformer (RoBERTa-style) pre-trained on 160M molecules, fine-tuned on BindingDB to learn potency-aware representations.
- Lean embeddings. Each molecule → 60-dim vector; SIMD-optimized cosine search at billion scale.
- Cloud-efficient. GPU indexing once; CPU search thereafter—cost-effective even for very large libraries.
- Generalizes. Works from ligand data alone—useful when 3D target structures are incomplete or unavailable.
Parkinson’s case study: LRRK2 (incl. G2019S)
- Query set: diverse known LRRK2 inhibitors (IC₅₀ ≤ 10 µM).
- Screen: retrieve 100k top candidates; prioritize CNS-like chemistry and novelty.
- Synthesis & assays:
- Hit ID: 87 tested → 4 with Kd ≤ 10 µM
- Expansion: 47 analogs → 10 additional actives (21%)
- Top hits: three sub-µM binders; improved affinity on wild-type LRRK2
- Outcome: rapid navigation to new chemical series suitable for lead optimization.
Scientific rigor
- Benchmarks: Performance competitive with state-of-the-art ML (e.g., Chemprop) across multiple targets.
- Open materials: Full Supporting Information (datasets, scripts, assay protocols) is available with the paper.
- License: Open access under CC-BY-NC-ND 4.0.
Why this matters for partners
- Speed to first chemistry: shorten hit ID cycles from months/years → weeks.
- IP-ready scaffolds: enforced novelty thresholds reduce prior-art collisions.
- Pragmatic scale: screen billions today; roadmap to trillion-scale spaces next.
Read the paper (JCIM, special issue) and explore the data. Ready to screen your target? Try BIOPTIC B1 and get a prioritized list of synthesize-ready candidates.