In Silico Approaches
The computational layer underneath every other Pharma program. State-of-the-art structural biology, generative chemistry, and ADMET prediction — but the differentiator is not the models, it is the closed loop. Every in silico prediction ships with a LabOS-orchestrated wet-lab validation plan, so customers stop paying for “predictions that don’t translate.”
The stack
Structure prediction
AlphaFold3, Boltz, RoseTTAFold-AllAtom for complex prediction. ESM3 for protein design. Full protein-ligand and protein-protein complex modeling with confidence calibration.
Generative chemistry
Diffusion-model-based de novo design. Goal-directed generation: optimize for binding affinity, ADMET profile, synthesizability, and IP novelty simultaneously.
Docking & virtual screening
GNINA, DiffDock, classical docking via Schrödinger Glide. Billion-scale virtual libraries (Enamine REAL) screened in days, not weeks. Top hits routed automatically into wet-lab confirmation.
ADMET & PK / PD
DeepChem-class models for absorption, distribution, metabolism, excretion, and toxicity. India-tuned PK models incorporating CYP2C9 / CYP2C19 / VKORC1 population-frequency adjustments.
Biomarker discovery
Hypothesis Agent over Apollo multi-omics — identify response-predictive and safety-predictive biomarkers from clinical-trial-ready cohorts.
MOA / off-target
Mechanism-of-action inference from transcriptomic perturbation data; off-target hit prediction via large-scale binding affinity models.
Why closed-loop matters
Pure in silico vendors (Schrödinger, ChemAxon, Cresset) deliver predictions; the customer is responsible for validating them. Our LabOS layer means every in silico claim ships with a typed wet-lab plan, and the results feed back into the model — so the second program from the same customer benefits from the first. This compounding is what software-only platforms cannot offer.
Engagement
$100K–$500K per focused project (single-target docking + biomarker discovery, ADMET panel, etc.). Or bundled into a Drug Discovery sprint at no additional cost.
Validation SLA
Every in silico deliverable ships with a typed wet-lab validation plan and scheduled CRO slot within 4 weeks. Predictions without validation paths are not billed as completed milestones.
Tooling matrix
| Task | Primary tools | Wet-lab confirmation |
|---|---|---|
| Structure | AlphaFold3, Boltz, ESM3 | SPR / cryo-EM partner slot |
| Virtual screen | GNINA, DiffDock, Glide | Biochemical assay at Syngene |
| De novo design | Diffusion generative models | Synthesis + IC50 panel |
| ADMET | DeepChem + India PK models | Microsomal stability, CYP panel |