Cutting-edge molecular simulation with no code
Gather data from across web databases (ChEMBL, BindingDB, and others). Dock the identified molecules in a simple conversation. Get SOTA binding scores and AI-predicted chemical properties in minutes.
Join our 51 'founding chemists' program for free Balto access, exclusive events, and rewards for usage milestones.
Gather data and model protein binding with a world-class docking algorithm in a simple AI-guided conversation.
Gather Data
Gather chemical data, convert PDF to SMILES, summarize and interact with articles.
- Database Access: Query ChEMBL, BindingDB, PDB, and other chemical databases directly
- PDF Extraction: Extract molecular structures and data from scientific papers
- Data Enrichment: Automatically annotate and enrich your compound libraries
Founding Chemist Program
Help us build the best forever-free AI assistant for drug discovery researchers. Free access, community, swag, and special events for early adopters.
| Feature | Included Monthly | Additional Usage |
|---|---|---|
| Monthly Subscription | Free | — |
| Core Tools | Free | — |
| Premium Tools | ||
| Docking | 30 dockings | $0.10/extra docking |
| Pocket Finder | 2 pose predictions | $5/extra pose prediction |
| PDF Analysis | 50 pages | $0.06/extra page |
| Solubility (LogS) | 30 predictions | $0.01/extra prediction |
| Lipophilicity (LogP) | 30 predictions | $0.01/extra prediction |
| Distribution (LogD) | 30 predictions | $0.01/extra prediction |
| hERG inhibition | 30 predictions | $0.01/extra prediction |
| CYP binding | 30 predictions | $0.01/extra prediction |
| Ames mutagenicity | 30 predictions | $0.01/extra prediction |
An AI Assistant is only as good as its tools
Enrichment in top 1% of returned hits
Enrichment Factor at 1% (EF1%) on DEKOIS 2.0
LogP Prediction (RMSE)
| Rank | Method | RMSE |
|---|---|---|
| 1 | OCHEM | 0.34 |
| 2 | Deep Origin | 0.449 |
| 3 | DNN (taut) | 0.47 |
| 4 | DNN (mono) | 0.5 |
| 5 | ACD/GALAS | 0.5 |
| 6 | ALOGPS | 0.5 |
| 7 | KOWWIN | 0.65 |
| 8 | JChem | 0.72 |
Ulrich N., Goss K. U., Ebert A. Communications Chemistry. 2021. Vol. 4(1). P. 90.
LogD Prediction (MAE)
| Rank | Method | MAE |
|---|---|---|
| 1 | Deep Origin | 0.425 |
| 2 | Chemprop - RDKit | 0.466 |
| 3 | Chemprop | 0.469 |
| 4 | BaseBoosting | 0.479 |
| 5 | ContextPred | 0.535 |
| 6 | GCN | 0.541 |
| 7 | AttrMasking | 0.547 |
| 8 | NeuralFP | 0.563 |
| 9 | AttentiveFP | 0.572 |
TDC.Lipophilicity_AstraZeneca
LogS Prediction (MAE)
| Rank | Method | MAE |
|---|---|---|
| 1 | Deep Origin | 0.525 |
| 2 | Chemprop-RDKit | 0.762 |
| 3 | AttentiveFP | 0.776 |
| 4 | Chemprop | 0.818 |
| 5 | RDKit2D + MLP | 0.827 |
| 6 | Basic ML | 0.828 |
| 7 | GCN | 0.907 |
| 8 | NeuralFP | 0.947 |
| 9 | CNN (DeepPurpose) | 1.023 |
TDC.Solubility_AqSolDB
Frequently Asked Questions
What algorithms does Balto use for molecular modeling?
Balto uses Deep Origin's proprietary algorithms for docking, properties predictions and finding novel pockets. Balto also relies on a few open-source tools and databases such as MolStar, PDB and ChEMBL.
How do you deal with hallucinations?
Although Balto hallucinations are very rare, it is advisable to verify the information that Balto provides.
What is included in the free trial?
Free trial includes 50 credits. These many credits are enough to: Dock and find poses for up to 410 molecules, Predict up to 910 molecular properties, Find novel pockets for up to 10 proteins, Analyze up to 305 pages of PDF documents. You can also explore countless free actions such as: Pose overlays and protein visualizations, Querying of integrated public databases.
Ready to accelerate your drug discovery?
Join thousands of researchers using Balto to streamline their workflows.