Deep Origin

What Deep Origin does

Deep Origin is an AI-powered drug discovery platform company headquartered in South San Francisco, founded in 2023 (originally as Formic Labs and rebranded in 2023; merged with BiosimAI in January 2024). The company builds physics-informed machine learning models that span quantum through cellular scales of biological organization, with the stated goal of replacing preclinical predictors that contribute to the approximately 90% clinical failure rate in drug development. Deep Origin primarily serves pharmaceutical and biotechnology companies engaged in target identification, hit discovery, lead optimization, program rescue, and trial design, alongside computational scientists and medicinal chemists who want self-serve access to molecular modeling tools.

The platform is accessed through three interfaces: Balto AI, a conversational natural-language assistant for drug discovery; DO Studio, a visual browser-based interface for docking, FEP, ADMET prediction, and structure preparation; and a Python API. Underlying capabilities include DODock (physics-informed ML docking with >90% PoseBusters success rate, 87.5% hit enrichment in top 1% on DEKOIS 2.0, and benchmark performance comparable to Isomorphic Labs' IsoDDE), FEP with Separated Topologies achieving sub-1 kcal/mol RMSE on 60% of targets with a fully automated pipeline, DO Patent (PDF-to-structure extraction at >98% accuracy, 12x faster than incumbents), and the Deep Origin Togo ADME models, ranked #1 on permeability, lipophilicity, solubility, half-life, hERG, and AMES tasks on Therapeutic Data Commons. The platform supports billion-molecule virtual screening via active learning, with up to 50 billion Enamine REAL Space compounds screenable in days.

Deep Origin operates three revenue streams: (1) subscription SaaS self-serve access to Balto AI and DO Patent with a freemium on-ramp of 30 free dockings per month and $0.10 per additional docking; (2) multi-year Discovery Partnerships offering full-stack platform access applied to customer targets with program milestones and optional IP co-ownership, with 2 of 40 partnership slots currently filled; and (3) out-licensing or co-development of internally validated programs, with GPR75 as the lead asset. The company is funded primarily through founder Michael Antonov's personal capital via Formic Ventures (proceeds from his Oculus exit) with approximately $50M raised, supplemented by a $31.7M ARPA-H CATALYST contract awarded in December 2025 to lead the PREDICTS consortium on drug safety modeling.

Name

Deep Origin

Legal name

Deep Origin Inc.

Website

https://deeporigin.com

Company type

Private

Founded year

2023

Operating status

Operating

Headcount range

101–250 employees

Short description

Deep Origin builds an AI-powered drug discovery platform combining physics-based modeling with machine learning to improve preclinical predictivity. It serves pharmaceutical companies via multi-year discovery partnerships, self-serve SaaS tools, and asset out-licensing including lead program GPR75.

Ownership category

akta.pro rank

Product category

AI Drug Discovery Platform

NAICS

Research and Development in Biotechnology (except Nanobiotechnology) (541714), Computer Systems Design and Related Services (54151)

SIC

Services-Commercial Physical & Biological Research (8731), Services-Prepackaged Software (7372)

akta.pro primary industry

AI/ML Platforms for Drug Discovery & Experiment Optimization (HLAGAJAO)

akta.pro secondary industries

Drug Discovery (Target ID/Validation & Hit Discovery) (HLAIALAA), In Silico / Computational Toxicology & PK/PD Modeling (HLAGAFAM)

Keywords

• Computational drug discovery
• AI drug discovery platform
• Molecular docking software
• Physics-based modeling
• Preclinical predictive models

Headquarters

HQ city

South San Francisco

HQ country

United States

HQ region

North America

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales, Infrastructure

Revenue model

1. SaaS Platform Subscription: Subscription-based access to Balto AI and DO Patent for teams with existing computational capability. Free trial on-ramp available. Self-serve model targeting computational scientists and medicinal chemists.
2. Discovery Partnership: Multi-year research collaborations providing full-stack access to Deep Origin's model suite applied to customer's specific drug target. Includes program milestones and optional IP co-ownership. 40 partnership slots available with 2 filled.
3. Asset Out-Licensing/Co-Development: Selected assets from internal validation pipeline available for out-licensing or co-development. Lead asset is GPR75. Provides upfront or milestone-based revenue for validated programs.

Pricing tiers

Go-to-market motion3 records

Distribution channels3 records

Marketing channels4 records

Core offering

Deep Origin builds an AI-powered drug discovery platform that combines physics-based modeling with machine learning to produce predictive preclinical models for pharmaceutical R&D. The platform is accessed via Balto AI (conversational assistant), DO Studio (visual interface), and a Python API, and provides tools for structure-based docking (DO Dock), binding affinity prediction (FEP), ADME property prediction (Deep Origin Togo), and PDF-to-structure extraction (DO Patent). The offering targets pharmaceutical and biotech companies seeking to reduce the ~90% clinical failure rate through better preclinical predictivity, delivered through SaaS subscriptions, multi-year research partnerships, and asset out-licensing.

Product overview

Deep Origin is an AI-powered drug discovery platform combining physics-based models with machine learning acceleration. The platform consists of three access interfaces: Balto AI (conversational AI assistant), DO Studio (visual interface), and Deep Origin API (code-first). Underlying capabilities include DO Patent (PDF molecular data extraction), DODock (structure-based docking and virtual screening), and FEP (free energy perturbation for binding affinity predictions). The platform targets pharmaceutical partners and research organizations seeking to reduce clinical trial failures through better preclinical predictive models.

Differentiator

Problem solved

Functional benefit

Brands

• Balto: AI assistant for drug discovery with conversational interface and molecular modeling capabilities
• DO Studio
• DO Patent
• DODock
• DO ABFE

Products and services

• Deep Origin Platform Unified AI-powered drug discovery workspace combining physics-based models with machine learning for predictive preclinical modeling across quantum, molecular, macromolecular, and cellular levels. Provides access to DO Dock, FEP, DO Patent, Balto AI, and DO Studio capabilities for pharmaceutical and biotech R&D teams.
• Balto AI Assistant Conversational AI assistant for drug discovery that enables molecular modeling, literature search, ChEMBL and PubChem queries, and docking through plain-English natural language interaction. Designed for medicinal chemists, biologists, and BD leads without command-line or DevOps requirements.
• DO Studio Browser-based visual interface for advanced drug discovery workflows, providing a unified UI for medicinal chemists to run docking, FEP, ADMET prediction, and structure preparation without command-line or DevOps.
• DO Patent Molecular data extraction tool that converts chemical structures from PDF documents into accurate, editable digital molecular structures. Reports >98% extraction accuracy and runs 12× faster than incumbent solutions.
• DODock (DO Dock) Physics-informed ML docking engine for pose prediction and virtual screening. Achieves >90% PoseBusters success rate, 87.5% hit enrichment in top 1% on DEKOIS 2.0, and retains predictivity at 0–20% Tanimoto similarity to the training set. Active learning enables screening of approximately 50 billion unenumerated Enamine REAL Space molecules in days.
• Free Energy Perturbation (FEP / DO ABFE) Production-grade binding affinity predictions using Separated Topologies (SepTop), enabling calculations on both congeneric series and structurally diverse molecules. Reports sub-1 kcal/mol RMSE on 60% of targets via a fully automated pipeline with zero manual pose curation.
• Deep Origin API Code-first programmatic access to the drug discovery platform. Python client with 50+ tools for pose prediction and billion-molecule virtual screening, with no DevOps required. Available via waitlist for early access.
• Discovery Partnership Multi-year research collaboration providing enterprise pharma customers full-stack access to Deep Origin's model suite applied to their specific drug targets. Includes program milestones and optional IP co-ownership, with 40 partnership slots available (2 currently filled).
• Validated Asset Out-Licensing / Co-Development Selected assets from Deep Origin's internal validation pipeline (lead asset: GPR75) made available for out-licensing or co-development with pharmaceutical partners. CD73 immuno-oncology program demonstrated a 30% hit rate against a 1% industry benchmark.

Quantifiable outcome

• 30% hit rate vs 1% industry benchmark on CD73 program
• +5 more outcomes

Named customers1 record

Segments2 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

AI capability8 records

Feature7 records

Partnerships

Five partnerships are on record, tiered core and minor.

• ArctoriscoreStrategic or Co-development Partner · 1 March 2026Collaboration funded by UK ARIA to discover a novel endometriosis therapy in less than nine months using AI and automated labs. The project aims to identify a drug target and validated molecule, significantly compressing traditional drug discovery timelines.
• Ginkgo BioworkscoreStrategic or Co-development Partner · 1 December 20254.5-year project funded by ARPA-H to develop a computational platform for drug safety. Deep Origin leads a team developing in-silico models to predict drug toxicity. Ginkgo contributes its 'Datapoints' platform to generate data supporting AI models.
• Tessel BiosciencescoreStrategic or Co-development Partner · 1 December 2025Tessel awarded up to $4.2M through ARPA-H CATALYST program as part of PREDICTS consortium led by Deep Origin. Tessel developing and scaling primary human tissue models of gut, kidney, and blood-brain barrier to measure drug movement through critical boundaries.
• NetriasminorStrategic or Co-development Partner · 1 December 2025Part of PREDICTS consortium led by Deep Origin under ARPA-H CATALYST program to develop computational platform for drug safety prediction.
• MIDOminorStrategic or Co-development Partner · 1 December 2025Part of PREDICTS consortium led by Deep Origin under ARPA-H CATALYST program to develop computational platform for drug safety prediction.

Scale indicators7 records

Recent moves6 records

Expansion highlights7 records

Direct peers

• Isomorphic Labs: Alphabet-backed AI drug discovery company whose IsoDDE (Drug Design Engine) was directly benchmarked against Deep Origin's DODock and demonstrated comparable performance. Both pursue physics-informed ML for structure-based drug design.
• Schrödinger, Inc. Public physics-based computational drug discovery platform with FEP+ and LiveDesign. Direct competitor in physics-based molecular modeling and enterprise pharma partnerships; considered the incumbent Deep Origin is positioning against.
• Recursion Pharmaceuticals: AI-driven drug discovery platform combining high-throughput biology with ML. Comparable in offering an end-to-end AI discovery stack, large pharma partnerships, and internal pipeline — and a named benchmark rival for Deep Origin Togo.
• Atomwise: AI small-molecule discovery platform using structure-based convolutional neural networks for virtual screening. Directly comparable product offering (docking, virtual screening) and pharma partnership motion.
• Insilico Medicine: AI drug discovery company combining generative chemistry, target discovery, and an internal clinical pipeline. Comparable in offering AI discovery plus asset out-licensing, and competing for similar enterprise pharma partnerships.
• XtalPi: AI + physics-based drug discovery platform combining quantum mechanics, molecular dynamics, and ML for crystal structure prediction and small-molecule discovery. Direct technology and customer overlap.
• Insitro: ML-driven drug discovery company combining genomics, phenotypic imaging, and computational chemistry for pharma partnerships. Overlapping platform thesis and partnership-driven revenue model.

Emerging players

• Iambic Therapeutics: Earlier-stage AI drug discovery company combining physics-based models with neural network potentials for predictive preclinical modeling. Direct technology overlap with Deep Origin's physics-informed ML approach.

Broad incumbents

• Google DeepMind (AlphaFold): Broad incumbent with AlphaFold 3 for protein structure and molecular prediction, explicitly benchmarked against DODock in Deep Origin's marketing. Represents both a technology rival and a potential partner/feeder for downstream platforms.

Others

• Ginkgo Bioworks: Synthetic biology and biological engineering platform currently partnering with Deep Origin on the ARPA-H PREDICTS consortium. Complementary rather than competitive; relevant for ecosystem positioning around AI + automated lab infrastructure.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks6 records

Key highlights7 records

Customer concentration

• GitHubgithub.com/deeporigin
• YouTubeyoutube.com/channel/UCX4vDqR9gD3bL_2pt9_4yWA
• LinkedInlinkedin.com/company/deeporigin
• Xx.com/DeepOriginBio

Revenue estimate

Valuation estimate

Number of profiles

Profiles12 records

Funding overview

Funding rounds1 record

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A1 record

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Deep Origin