Deep Origin API

The Deep Origin API is a unified programmatic interface designed for drug discovery researchers and computational chemists who want to build end-to-end molecular workflows without any configuration or DevOps overhead. By consolidating state-of-the-art molecular modelling, cheminformatics, docking, ADMET profiling, and free energy calculations into a single API, Deep Origin enables teams to go from data retrieval to actionable results with minimal lines of code.

The platform is aimed at scientists and software developers in the life sciences who need rapid, scalable access to best-in-class computational tools — from pocket identification and molecular docking through to toxicity prediction and retrosynthetic analysis — all accessible programmatically and ready for integration into custom pipelines.

Core Docking Workflow Capabilities

• Pocket finding and comparison: Source chemical data from across the web and identify binding pockets with minimal code. Pocket data can be generated based on crystal structure, novelty, XYZ coordinates, or residue IDs. Results include Pocket ID, Drugability Score, Volume, Total SASA, Polar SASA, Polar/Apolar SASA Ratio, Hydrophobicity, and Polarity.
• Docking grid box visualisation: Validate pocket boundaries and grid box dimensions for docking in a single function call, with built-in support for mismatched or manually adjusted inputs using either precomputed or custom protein data.
• Docking scores, poses, and reporting: Run molecular docking with a single function call using preprocessed or custom inputs. Generate poses, binding energies, ranking scores, and interactive visualisations in seconds, ready for analysis, filtering, or export.
• ADMET prediction and export: Predict absorption, distribution, metabolism, excretion, and toxicity properties for docked ligands, ready for integration into downstream pipelines or tools of your choice.

Molecular Data Retrieval and Informatics

• Database integrations: Direct access to Protein Data Bank (PDB), ChEMBL, UniProt, PubChem, BindingDB, Zinc DB, Mcule Express, AlphaFold DB, and a Unified DB that locally integrates ChEMBL and BindingDB data.
• Intellectual property and literature tools: Patent checking via SMILES input, automated scientific literature search, PDF document analysis for extracting and summarising research findings, and conversion between molecular identifiers including SMILES, CAS, and IUPAC formats.
• Additional informatics tools: General web search for drug discovery information and Draco 1.0, a tool that converts PDF documents into SMILES representations.

Protein Structure Analysis and Ligand Preparation

• Protein processing: Binding pocket identification to detect potential binding sites, and protein structure alignment.
• Ligand tools: Protonation state prediction at a given pH, functional group identification for molecular substructures, and molecular similarity search to find analogues to a given input.

Molecular Dynamics and Free Energy Calculations

• RBFE: Computes relative binding free energies with high accuracy.
• ABFE: Calculates absolute binding free energies of ligands to proteins.
• Simple MD: Fine-grained molecular dynamics simulation for structure relaxation and analysis.

Structural and Visualisation Tools

• Rendering and visualisation: 2D molecular diagram generation from SMILES, 3D protein visualisation, binding pocket highlighting, and a docking result viewer for inspecting ligand poses and interactions.
• Structural analysis: RMSD calculation for comparing molecular conformations and structural changes, and mutagenesis analysis for identifying mutations or structural gaps.

Molecular Property and ADMET Profiling

• Drug-likeness descriptors: Synthetic Accessibility Score (SAS) to assess molecular complexity, and Quantitative Estimate of Drug-likeness (QED).
• Physicochemical properties: Predictions for LogP (lipophilicity), LogS (solubility), LogD (distribution coefficient accounting for ionisation), and pKa (acidity values).
• Toxicity predictions: hERG inhibition and binding prediction for cardiotoxicity risk assessment, CYP450 interaction and binding prediction for metabolic stability, and Ames mutagenicity assessment for mutagenic and carcinogenic potential.
• Multi-parameter optimisation: A scoring engine (Togo) for compound scoring across multiple parameters simultaneously.

Molecular Docking Tools

• Docking simulation and analysis: Standard molecular docking simulations, binding energy and interaction analysis, 3D docking box visualisation, and ligand and protein preparation utilities.

Data Processing, Batch Pipelines, and Workflow Automation

• CSV dataset handling: Read, clean, and transform molecular datasets, with SQL-like querying capabilities for filtering CSV data.
• Batch processing: Efficient handling of large molecule datasets and simultaneous batch docking of multiple molecules.
• Workflow automation: File management for listing and retrieving files from a working directory, and a basic retrosynthetic engine for predicting simple synthetic routes.

The Deep Origin API is designed to require no configuration or DevOps setup, making advanced computational chemistry capabilities immediately accessible to development teams. Deep Origin is headquartered in South San Francisco, CA, and the API is currently in a coming-soon phase with launch notifications available for interested users.