GINKGO DATAPOINTS · BOSTON · PART OF GINKGO BIOWORKS
We define the category of Bio × AI data provider.
Trusted by top pharma, biotech, techbio, and pure tech companies. Billions of biological data points generated using bonafide validated methods. Run by robots, not by hand. Expert human consultation always included.
[ Disclosed Partners & Customers ]
[ Data Generation Services ]
Three platforms.
Countless use cases.
One data generation partner.
AI/ML underscores everything — or not. You decide.
01
Antibody Developability
Biophysical characterization for drug-like antibodies.
Screen over 2,400 antibody designs in parallel across 12+ biophysical assays. AI-ready data with run-to-run consistency validated against published literature — from pilot screens to full training sets.
View full service details →02
Functional Biology
Perturbation response profiling and specialized HTS in one platform.
The world leader in DRUG-seq. Industrialized DRUG-seq and Cell Painting at library scale across 80+ cell types, including primary cells — plus custom high-throughput screens built around your target and biology. From target ID to hit-to-lead, with AI/ML-ready output.
View full service details →03
Small Molecule ADME
ADMET testing at lead or library scale — now available as ADME-One™.
ADME-One™ — our integrated Tier 1 ADME offering with Inductive Bio (AI human PK projection) and Tangible Scientific (compound management). Single US-based workflow, rapid results, pricing that beats offshore. Individual à la carte assays remain available, with structured ML-ready data returns.
View full service details →[ DATA, DIFFERENTLY ]
Data as infrastructure.
We treat biological data the way software companies treat code — versioned, validated, and built to scale.
Traditional
Filter compounds early, characterize comprehensively later. Fast to start, but deferred risk compounds downstream failures.
Integrated
Screen entire libraries — not just leads — with multi-modal assays run in parallel. Reduces decision risk at every stage.
Foundational
Generate the vast, diverse datasets that train predictive AI/ML systems. Accelerate downstream and increase probability of success.
[ Signal Work ]
Where we're putting
the field on notice.
We don't just generate data, we set standards. From landmark government contracts to open-science consortia, this is what category creation looks like.
ARPA-H CATALYST
FDA-qualifiable in silico ADME-Tox prediction platform
Teaming with Deep Origin and partners to build a predictive toxicology platform that will screen 45+ primary cell types across 5 tissue types — liver, intestine, blood, bone marrow, kidney — using DRUG-seq, High Content Imaging, and Metabolomics.
[ Flagship Open Science ]
Virtual Cell Pharmacology Initiative
Building the world's highest fidelity public gene-drug response atlas. 100,000 compounds. 6-point dose-response. All measured with high SNR DRUG-seq. Free access to data, plus free data for your compounds (open submissions are now live).
thevirtualcell.com →Antibody Developability Consortium
Now launched with founding members AbbVie, argenx, Lundbeck, and Takeda, and still open to new members. Together we're building the field's largest standardized developability dataset: 10,000 antibodies drawn from member and public sequences. Ginkgo trains a foundation model on it, and members fine-tune that model on their own molecules through Apheris' federated infrastructure, so no raw sequences are exposed. The first data reaches members in early 2027.
Read the announcement →AbDev Competition
The first open competition to benchmark antibody developability modeling, measuring the state of the art and establishing evaluation standards for the field using PROPHET-Ab data. Hosted on Hugging Face. Published in mAbs journal.
View results →Datapoints Summit
Annual summit convening the field's leading voices in Bio × AI. Past speakers include OpenAI, Google, Novartis, Microsoft, NVIDIA, Pfizer, Amgen, Bioptimus, Broad Institute, MD Anderson, and Recursion.
BOSTON Open Science Forum
Gathering of Boston-area researchers committed to open data, reproducible science, and collaborative discovery.
[ Collaborators & Clients ]
Who we build with.
Who we build for.
Some relationships are about integration: shared infra, closed loops, models that improve with each run. Others are about delivery: rigorous data, on time, that you own outright.
Inductive Bio's Compass platform lets chemists design and rank small molecules using ADMET prediction models that placed first in both the Polaris and OpenADMET blind challenges. This is the AI layer for ADME-One™.
Natick, MA — Tangible handles secure compound storage and routes samples to our ADME service via API-enabled order submission. Researchers get structured, metadata-rich results in days. This is the logistics layer for ADME-One™.
Apheris powers the federated infrastructure behind the Antibody Developability Consortium, now launched with founding members AbbVie, argenx, Lundbeck, and Takeda. Ginkgo generates the 10,000-antibody dataset and trains a foundation model on it. Apheris delivers that model into each member's own environment, where they fine-tune it on their proprietary molecules without exposing their IP.
[ Open Datasets — HuggingFace ]
Public data that the
ML community actually uses.
Datasets open for the community. Available here and also on HuggingFace. This is what "AI-native" looks like in practice.
[ Get Started ]
Your AI models need
better biological data.
Fee-for-service. You own the data. No royalties. Scale from pilot to full model training — we're ready when you are.












