Axiom Bio
Its own human liver-cell lab plus a model that predicts which drugs will injure the liver, pitched as a replacement for animal toxicity studies.
Neil Gaikwad · axi.om
The bet in one paragraph
Liver injury is one of the most common reasons drugs fail late or get pulled, and the animal studies meant to catch it miss many human cases. Axiom screens compounds on primary human liver cells in its own lab, links those readings to what happened to real patients, and sells the prediction. The FDA's 2025 plan to move away from animal testing creates a buyer for exactly this evidence. If Axiom's data proves predictive in blinded tests with pharma, it is building the human-relevant toxicity dataset that a regulator eventually accepts.
Why this team
CEO Brandon White built machine learning platforms at Uber and led products at Freenome and Spring Discovery. CTO Alex Beatson has a Princeton PhD in machine learning and worked on molecular models at Genesis Therapeutics and Redesign Science. Both are co-authors on a 2026 Cell Systems paper on imaging liver cells.
The open question is biology leadership. The founding biologist who built the liver-cell assays left in May 2026, and the wet lab is the moat.
- Brandon WhiteCo-founder and CEOMachine learning engineer at Uber from 2014; early at Freenome; about four years leading product at Spring Discovery. research.contrary.com
- Alex BeatsonCo-founder and CTOPrinceton PhD in machine learning; Google; molecular modeling at Genesis Therapeutics and Redesign Science. research.contrary.com
What exists today, and what does not yet
Today
A wet lab screening molecules on primary human liver cells with imaging and mass spectrometry, and a product called Translate that predicts liver injury risk, its likely mechanism and a safe dose. As of June 2025 it reported 130,000+ molecules screened and 394 million labeled cells.
Not yet
Any organ beyond the liver, a path to FDA qualification, and a published result from a blinded pharma study. The company's own documents give sensitivity anywhere from 51% to 75% depending on the test set.
Traction on the public record
- Pilots underway or being finalized with six of the top 20 pharma companies, none named. Apr 2025 · biopharmatrend.com
- Released a public dataset explorer covering 130,000+ compounds screened on human liver cells. Jun 2025 · biopharmatrend.com
- No named customer, revenue or blinded study outcome disclosed. Sep 2026 · research.contrary.com
- Eight open roles, including its first commercial hire and a medicinal chemist. Aug 2026 · api.ashbyhq.com
What I cannot see from outsideWhether the pharma pilots converted, at what contract size, and what the blinded results showed. CRV and Abstract are listed as investors on Axiom's site without an announced round, which may mean the company has already raised again.
Competition, and why they could still win
| Company | Stage | How it differs |
|---|---|---|
| Absentia Labs | Pre-seed | Its AI liver model is the first in silico tool accepted into the FDA's qualification program, step one of three. fda.gov |
| Inductive Bio | Series A, $25M, plus ARPA-H funding | Broader drug-property prediction, now generating its own toxicity data with Amgen. pharmaceutical-technology.com |
| Revalia Bio | Seed, $14.5M | Tests drugs on donated human organs: higher fidelity, lower throughput. biospace.com |
| Contract research organizations | Incumbents | Own today's toxicity budget at $3,000 to $15,000 per compound. research.contrary.com |
A model alone is easy to copy. A wet lab that has screened more than 130,000 compounds on consistent human donor cells is not. If Axiom can show its predictions catch human liver injury that animal studies miss, its data becomes the reference set for the whole market, including for regulators.
This company sits inside my big bet Medicine becomes a design problem.
What I would have to believe
- Blinded pharma studies confirm the model catches human liver injury better than current assays.
- Pharma will pay platform prices for prediction instead of per-assay prices for lab work.
- Regulatory acceptance follows the data, even if a competitor entered FDA qualification first.
Questions for the first call
- What did the blinded pharma studies show, and how many pilots converted to paid contracts?
- Why does sensitivity range from 51% to 75% across your own reports, and which number should a buyer trust?
- Who leads biology now, and how do you protect assay quality after the founding biologist's departure?
- Are you pursuing FDA qualification, and how do you respond to Absentia's acceptance into the program?
- Have you raised since the seed, and on what terms?
- When does a second organ ship, and which one?
What gets them to a Series A
- Named pharma customers on multi-year contracts.
- Published results from at least one blinded prospective study.
- A second organ model in customers' hands.
- A documented regulatory strategy with the FDA.
When I would walk away
- Blinded studies show no clear advantage over existing assays.
- Pharma keeps paying for pilots but will not sign platform contracts.
Risks
- A rival has the first FDA acceptance for an AI liver-injury model. fda.gov
- Inconsistent accuracy figures across the company's own documents. scalingbiotech.substack.com
- No named customer or revenue 18 months after the pilots were reported. research.contrary.com
- The founding biologist behind the core assays left in May 2026. research.contrary.com
Sources
- Amplify Partners, Our investment in Axiom (Apr 2025)
- Fenwick, Axiom $15M Series Seed
- BioPharmaTrend, Axiom seed funding (Apr 2025)
- BioPharmaTrend, Axiom liver toxicity dataset (Jun 2025)
- Contrary Research, Axiom (Sep 2026)
- Scaling Biotech, Can an AI beat an in vivo model?
- FDA, First in silico tool accepted under ISTAND
- Axiom
Written from public information only. No contact with the company, and no position in it. Nothing here is investment advice: these are my own notes on public information, written to sharpen judgment, not to recommend buying or selling anything.