Eliminating drug toxicity with AI: our investment in Axiom

An Unsolved Problem That Costs Billions

One of the largest problems in drug development is toxicity - the presence of unintended adverse effects on the human body. Hepatotoxicity (liver toxicity) alone accounts for almost 25% of development-stage drug failures and nearly a third of market withdrawals of approved drugs. The financial impact of those failures is staggering, with billions of dollars worth of R&D investment – and tens of billions of dollars of pharma market cap – lost annually due to liver tox failures. Drugs that will ultimately fail are advancing far too long in the development process, consuming scarce dollars, time, and research capacity. These failures make successful therapies more expensive and slower to develop, which directly harms patients.  

Unfortunately the industry is barely making progress on the problem. Researchers are basically using the same tools and techniques as they were decades ago, keeping development costs and failure rates still stubbornly high.The current standard practice for testing remains the animal study, which in many cases is no better than a coin-flip at predicting if a drug will ultimately be toxic to humans. The entire process remains antiquated, expensive, and not very effective. In fact, it's bordering on broken.

We believe Axiom is the most promising company yet to tackle this problem. In a short time, they have built the world’s largest dataset dedicated to chemical toxicity in human cells, trained the industry’s best performing AI models for liver toxicity prediction, and are already delivering massive value to biotech and pharma customers.

Today, Amplify is thrilled to announce our seed investment in Axiom, and our partnership with its two founders, Brandon White and Alex Beatson. Since our initial investment in the company in 2023, Brandon, Alex and the entire Axiom team have made astonishing progress in accurately predicting drug toxicity through in silico modeling and are poised to transform the industry’s entire approach to drug safety testing. 

A Computational Approach, Rooted in Experimental Biology

Axiom is advancing the frontier of drug toxicity testing. Their vision of the future is not based on animal biology that fails to translate to human outcomes, or small pockets of third-party data assembled to make incrementally better tox predictions. Instead, Axiom believes that if we test all types of human cells against a never-ending stream of chemical compounds, and then accurately measure the biological effects of those experiments, we can create the definitive dataset on human cell toxicity. That dataset would serve as the atlas of chemical-cell interactions, but more importantly, would form the foundation for AI models capable of accurately predicting the toxicity of any novel chemical compound imagined by humans or computers. 

This approach is not theoretical; Axiom has made it reality. In eighteen months since founding, Axiom has built a proprietary biological experimentation platform that screens chemical compounds against live human hepatocytes (liver cells) at massive scale. Then, using advanced imaging techniques, the company captures hundreds of unique data points per cellular experiment. This process has been repeated across tens of thousands of molecules and millions of human cells. This dataset, which continues to expand at an accelerating rate, currently consists of:

• 115,000+ unique small molecules
• 394+ million individually labeled cells
• 9+ billion individually labeled mitochondria
• 5,200+ clinical drug-induced liver injury (DILI) likelihood assessments
• 38,400+ liver enzyme elevation datapoints

Axiom’s world-leading dataset now powers their proprietary AI models, which have already proven they can outperform traditional experimental methods in predicting liver toxicity - across sensitivity, specificity, and AUC - at substantially less cost and time.

But what will these advances mean to customers?  We believe that Axiom’s combination of proprietary biological experimentation, and predictive AI, all delivered through a simple software interface, will fundamentally reshape how the biopharma industry perceives tox testing. And with outstanding results in-hand for hepatocyte models, we’re confident that Axiom can expand their suite of predictive models to include toxicity in the heart, kidney, brain, skin and other key organs.

More important than being “better, faster, and cheaper” tox testing, we believe Axiom is building the most comprehensive AI model of whole-body human toxicity, which will allow the industry to accurately understand the toxicological effects of 100x to 1000x more compounds while simultaneously reducing development costs and accelerating development timelines. 

We’re also convinced that Axiom will allow the industry to pull tox testing forward in their drug development process. When presented with two to three orders of magnitude improvements in cost and time - coupled with more accurate tox predictions - we think tox testing will be one of the first steps taken in every drug development process. As Axiom’s models expand and improve, they will provide researchers with accurate clinical risk assessment at human-relevant doses, clearer understanding of feature importance, and a better interpretation of underlying biological mechanisms - all of which is well beyond the scope of what the industry can achieve today. 

The Right Solution at the Right Time

The need for pharma to adopt solutions like Axiom has taken on a new urgency. Just this month the FDA announced a mandate to phase out certain animal testing requirements, with FDA Commissioner Makary specifically calling out “AI-based computational models of toxicity” as tools that can increase drug safety, improve development timelines, and lower cost to consumers.

Building on their standout showing at the Society of Toxicology conference in March, we’re confident that Axiom will soon become the face of the in silico testing market.

World-Class Founders with Deep Domain Expertise

Brandon White (CEO) and Alex Beatson (CTO) bring exceptional talent to this challenge. Brandon and Alex are members of the exclusive, but growing, ranks of machine learning engineers who’ve dedicated their careers to the life sciences. With published research at the intersection of machine learning and computational biology, and experience building tools at companies like Freenome, Spring Discovery, Genesis Therapeutics and Redesign Sciences, they are the right people to bring a computational approach to revolutionize toxicology and drug development. 

Funding, and More Funding

We were lucky to have known Alex and Brandon individually for years, and were thrilled to learn that they had teamed up to tackle this problem. Amplify is proud to have led Axiom’s $7 million seed round in 2023, and eagerly invested again alongside our friends Dimension Capital in the company’s $8 million financing late last year. We’re joined by an excellent syndicate that includes Zetta Venture Partners and angel investors including Jeff Dean, Laksh Aithani, Barry McCardel, Alec Nielsen, Ari Morcos, Stef van Grieken, Elliot Hershberg, and others.

We believe Axiom is positioned to transform the multi billion market for drug toxicity testing. Their data-first approach will change how drug developers do their jobs and will ultimately get safer drugs to patients faster. With $15 million of funding, growing validation from customers and partners, and an ambitious vision to build a comprehensive model of human cell toxicity, we couldn't be more excited for the journey ahead.

‍