CarbonAI® is a de novo small-molecule and PROTAC AI design engine capable of screening billions of compounds in mere days, simultaneously optimizing multiple pharmacological properties for lead generation and optimization. It delivers 50 candidates in 3 iterations for wet-lab testing, with milestone success achieved.
Scaffold Hopping de novo Generation
Lead Optimization Analog Expansion
PROTAC de novo Generation
ADME(Absorption, Distribution, Metabolism, Excretion) Prediction
Toxicity Prediction
DMPK Prediction
Protein Binding Specificity Optimization
Target Binding Pocket Detection
Molecular Docking
Compound Off Target Toxicity Prediction
| Design approach | How it works | When to choose it |
|---|---|---|
| Scaffold-hopping de novo lead generation |
Explores 10⁷–10⁹ molecules and proposes 10–100 novel chemotypes. No 3-D structure required. |
When you have no starting hits or want new IP space. |
| Lead-optimization / analog expansion | AI refines existing hits for potency, selectivity, ADME/Tox, and developability. | When you already have hits but need a better profile. |
| PROTAC de novo design | Builds bifunctional degraders and ranks them for ternary-complex stability. | When you aim to degrade a disease-relevant protein. |
| Target-binding & off-target optimization | Predicts binding and off-target liability; performs rapid docking. | Use throughout the project for balanced efficacy and safety. |
*Timelines can overlap to shorten total project duration.
| Item | What’s included | Batch size | Timeline* |
|---|---|---|---|
| Computational design | De novo or lead-optimization; ranked list with predicted potency, selectivity, ADME/Tox. | — | ~1 week |
| Compound synthesis | Route design, synthesis ≥ 95% purity, analytical QC. | Up to 10 compounds | ~2 weeks |
| ADME / Tox profiling | hERG, CYP panel, solubility, stability, permeability + wet-lab confirmation. | Matches set | 1 week |
| Biochemical / cell assays | IC₅₀/Kᵢ or cellular potency; off-target panel if requested. | Matches set | 1–2 weeks |
*Runs overlap whenever feasible.
No. CarbonAI® can work from sequence alone, learning pocket flexibility directly from primary sequence. A high‑resolution structure improves docking accuracy but is not mandatory.
No. Ainnocence’s discovery platform is a proprietary, sequence-based AI system developed entirely in-house. By working directly from primary sequence data, it eliminates the need for external 3-D structure-prediction software and other open-source packages. Instead, our technology performs ultra-high-throughput virtual screening and multi-objective optimisation, evaluating up to 10¹⁰ candidates within hours to days—giving our partners capabilities that off-the-shelf tools cannot match.
