Wet AI™

Your found them. We will tell you which ones work.

High-throughput validation of AI-designed compounds in the most genetically diverse human cell models available. Weeks, not quarters.

Available now

Wet AI™

Your found them. We will tell you which ones work.

High-throughput validation of AI-designed compounds in the most genetically diverse human cell models available. Weeks, not quarters.

Available now

Early Prioritisation

Choose leads early on, reduce late stage failure.

Human data

We test safety, efficacy and global response.

1000x

Automation enable screening at scale, test thousands of drug simultaneously.

≥5M

viable cells per vial

Become smarter

Access our data lochs and to make your AI smarter.

The bottleneck moved

Generating molecules stopped being the hard part. Choosing between them didn't.

Your platform proposes more credible candidates than any preclinical budget can prosecute. The models ranking them were trained on single-donor, immortalised, historically narrow biology.

A model is only as representative as the biology underneath it. Wet AI puts real human biology under the decision, at the speed the candidates arrive.

The bottleneck moved

Generating molecules stopped being the hard part. Choosing between them didn't.

Your platform proposes more credible candidates than any preclinical budget can prosecute. The models ranking them were trained on single-donor, immortalised, historically narrow biology.

A model is only as representative as the biology underneath it. Wet AI puts real human biology under the decision, at the speed the candidates arrive.

How it works

Breadth first, depth last. Cheap failures happen early; the expensive experiments only run on compounds that earned them.

Wet AI™

Your found them. We will tell you which ones work.

High-throughput validation of AI-designed compounds in the most genetically diverse human cell models available. Weeks, not quarters.

Available now

Compounds reaching the multicellular models arrive with a mechanistic and donor stratified case already built.

Why it works

Diveristy

Eighteen genomes, not one

CYP3A4 activity spans fifty-fold across the panel. Clean in one background is not clean. Every line HLA-typed, three ancestries, 1:1 female to male.

Depth

Phenotype, not one endpoint

Six-channel Cell Painting reads hundreds of features per cell. Separates mechanisms of toxicity instead of answering one predefined question.

Human

Multicellular, one genome

Hepatocytes with Kupffer, stellate and endothelial cells from the same donor. No allogeneic noise confounding the compound effect

CYP3A4 activity across eighteen Cytochroma iPSC donor lines, spanning a greater than fifty-fold range..

Proof, not promise

MASH, peer-reviewed

Lipid-loaded Cytochroma hepatocytes discriminated target-selective Cyclophilin B inhibitors from Cyclophilin A, at potency comparable to FDA-approved resmetirom. J Med Chem 2025, 68(6), 6815 (DOI).

Benchmarked on automation

97.9% post-thaw viability. Multiplexed Cell Painting resolving distinct mechanisms of toxicity across an eight-compound panel. Automated seeding roughly halved well-to-well variability.

Consortium-grade

Advanced Liver Partner in the OASIS Consortium, convened by HESI Global and the Broad Institute of Harvard and MIT. Tox Sci 2025, 208(2), 225 (DOI).

Cardiac

Ventricular cardiomyocytes with quantified contractile activity and confirmed sarcomeric organisation, across the same diverse panel.

Where the regulators are heading

The FDA published its Roadmap to Reducing Animal Testing in Preclinical Safety Studies in April 2025, and reported hitting year-one goals in April 2026 — including draft guidance on weight-of-evidence approaches that names in vitro assays and human-relevant models directly.

Our models are New Approach Methodologies of exactly that kind. Screening across a population-representative human panel is not a workaround for the animal study; it answers a question the animal study cannot.

Data Lochs

Deep reference datasets, built on our models with commercially available compounds.

Labelled, human-relevant, donor-stratified data at training scale is what computational teams consistently find missing. Data Lochs are that substrate.

They sit on the same models your compounds would be screened in, so a new screen lands in a coordinate system that already exists.

Proof, not promise

MASH, peer-reviewed

Lipid-loaded Cytochroma hepatocytes discriminated target-selective Cyclophilin B inhibitors from Cyclophilin A, at potency comparable to FDA-approved resmetirom. J Med Chem 2025, 68(6), 6815 (DOI).

Benchmarked on automation

97.9% post-thaw viability. Multiplexed Cell Painting resolving distinct mechanisms of toxicity across an eight-compound panel. Automated seeding roughly halved well-to-well variability.

Consortium-grade

Advanced Liver Partner in the OASIS Consortium, convened by HESI Global and the Broad Institute of Harvard and MIT. Tox Sci 2025, 208(2), 225 (DOI).

Cardiac

Ventricular cardiomyocytes with quantified contractile activity and confirmed sarcomeric organisation, across the same diverse panel.

Where the regulators are heading

Data Lochs

Deep reference datasets, built on our models with commercially available compounds.

Labelled, human-relevant, donor-stratified data at training scale is what computational teams consistently find missing. Data Lochs are that substrate.

They sit on the same models your compounds would be screened in, so a new screen lands in a coordinate system that already exists.

50 μM, 48 h · n=3 per condition

Applications

Drug-induced liver injury (DILI) and hepatotoxicity screening
MAFLD and MASH disease modelling with quantifiable steatosis by high-content imaging
Cytochrome

Multi-donor CYP450 profiling

Diverse-donor iHeps express key CYP450 enzymes with functionally relevant induction and reproducible donor-dependent variability, unlike HepG2 or short-lived primary hepatocytes.

Consortium participation

Cytochroma is the Advanced Liver Partner to the HESI OASIS consortium, convened by HESI Global and the Broad Institute of Harvard and MIT, contributing diverse-donor human liver models to its hepatic-safety programme.

“Their hepatocyte MASH models helped us quickly evaluate novel inhibitors in a biologically relevant model, essential data for our J. Med. Chem. publication.”

Dr Dahlia Doughty-Shenton, University of Edinburgh

OASIS Consortium

Cytochroma is an Advanced Liver Partner in the OASIS Consortium (’Omics for Assessing Signatures for Integrated Safety), a pre-competitive hepatic-safety consortium convened in July 2023 by HESI Global and the Broad Institute of Harvard and MIT as a working group of HESI’s eSTAR committee. OASIS integrates transcriptomics, proteomics and high-content Cell Painting across human-relevant models, with an initial focus on liver safety, and spans more than 160 experts across 17 academic institutes, 7 government agencies, 17 industry organisations and 3 NGOs. Cytochroma contributes diverse-donor human liver models to the consortium’s hepatic-safety programme. See The OASIS Consortium: Integrating Multi-Omics Technologies to Transform Chemical Safety Assessment, Rouquié et al., Toxicological Sciences 2025, 208(2), 225 (DOI 10.1093/toxsci/kfaf128).

Frequently asked questions

What is Wet AI?

Why do AI-generated compounds need wet-lab validation?

How fast is a Wet AI screen?

Which organ systems can be screened?

How does donor diversity change the result?

What is Cell Painting and why does it matter for prioritisation?

How is the data delivered?

What are Data Lochs?

What evidence is there that these models predict clinical outcomes?

How does this relate to FDA guidance on New Approach Methodologies?

Who runs the screens?

Where is the work carried out?

Request a quote or technical discussion

Tell us your assay format, donor requirements, and timelines. Our scientific team will help select the right donor panel and model configuration.

Contact Cytochroma