Balto AI

Balto is an AI-powered drug discovery assistant developed by Deep Origin that enables medicinal and computational chemists to perform cutting-edge molecular simulation through a simple, code-free conversational interface. Designed to make world-class docking and cheminformatics tools accessible without requiring deep technical expertise, Balto allows researchers to gather chemical data, identify binding pockets, and dock small molecules in minutes — all within a guided AI conversation.

The platform is built for drug discovery researchers who want immediate access to state-of-the-art molecular modeling capabilities without waiting on technical teams, lengthy procurement processes, or paying for unused functionality. Balto operates on a flexible, usage-based pricing model with a free tier available, making it suitable for individual researchers as well as larger organizations.

Core Workflow

1. Gather Data: Retrieve chemical and biological data from major public databases including ChEMBL, BindingDB, PDB, PubChem, and UniProt. Convert PDF documents to SMILES structures, summarize scientific articles, and interact with literature directly within the platform.
2. Find Pockets: Locate binding pockets on target proteins and strategize your investigation guided by a chemistry-aware AI assistant.
3. Dock Small Molecules: Run benchmark-topping molecular docking using Deep Origin's proprietary algorithms, obtaining state-of-the-art binding scores through a conversational interface requiring no coding.

Key Features and Capabilities

• Conversational AI assistant with life science and cheminformatics awareness
• World-class molecular docking benchmarked on PDBBind 285 and DEKOIS 2.0 datasets
• Novel binding pocket identification and pose prediction
• PDF analysis with structure-aware parsing to extract molecules from literature and patents
• AI-predicted chemical properties including solubility (LogS), lipophilicity (LogP), distribution (LogD), hERG inhibition, CYP binding, and Ames mutagenicity
• Protein and pose visualization, including pose overlays
• Querying of integrated public databases at no additional cost
• Retrosynthetic AI and molecular dynamics capabilities coming soon

Benchmark Performance

• LogP prediction: Deep Origin ranks 2nd globally with an RMSE of 0.449, outperforming models such as DNN, ACD/GALAS, ALOGPS, and JChem
• LogD prediction: Deep Origin ranks 1st with an MAE of 0.425, ahead of Chemprop-RDKit and other leading models
• LogS prediction: Deep Origin ranks 1st with an MAE of 0.525, significantly outperforming Chemprop-RDKit, AttentiveFP, and other benchmarked approaches

Pricing Tiers

• Basic (Free with quota): Includes molecular modeling and drug discovery assistant, benchmark-topping docking and pocket-finding tools, and 3 organization seats. Free quota includes 30 dockings, 2 pocket finder pose predictions, 50 pages of PDF analysis, and 30 predictions each for LogS, LogP, LogD, hERG inhibition, CYP binding, and Ames mutagenicity. Additional usage is available at low per-unit rates.
• Professional (Coming soon): Everything in Basic plus increased batch sizes for docking, 20 organization seats, and free energy perturbation (FEP) capabilities.
• Enterprise (Coming soon): Everything in Professional plus unlimited organization seats, discounted credits, generative AI for molecules, and ultra-large virtual screening.

API and Code-Based Access

• Query major chemical and biological databases including PDB, ChEMBL, PubChem, and UniProt programmatically
• Automate structure retrieval, annotations, and internal data enrichment
• Extract molecules from literature and patents, run similarity searches, and analyze PDFs with structure-aware parsing
• Run high-throughput docking and property calculations using state-of-the-art AI tools at scale
• Optimize lead compounds with code-based precision for computational chemistry workflows

Balto is developed by Deep Origin, headquartered in South San Francisco, CA. The platform relies on Deep Origin's proprietary algorithms for docking, property prediction, and pocket finding, and also integrates open-source tools and databases such as MolStar, PDB, and ChEMBL. An early adopter "Founding Chemists" program is available, offering free access, community events, and rewards for usage milestones.