Liver Injury

Axiom accurately predicts human liver toxicity

20-25% of clinical trials fail due to drug-induced liver injury, Axiom wants to eliminate these failures

SOURCE “Setting Clinical Exposure Levels of Concern for Drug Induced Liver Injury (DILI) Using Mechanistic In Vitro Assays” F Shah et al, Toxicological Sciences, 2015

Pfizer

2D HepG2

Axiom

AI model

ROC-AUC--0.89
Sensitivity / True Positive Rate34%75%
Specificity / True Negative Rate91%90%
Cost$3,000–$5,000$100–$450
SOURCE “Setting Clinical Exposure Levels of Concern for Drug Induced Liver Injury (DILI) Using Mechanistic In Vitro Assays” F Shah et al, Toxicological Sciences, 2015
SOURCE "Evaluation of the use of imaging parameters for the detection of compound-induced hepatotoxicity." S Schadt et al, Toxicol In Vitro, 2015

AstraZeneca

2D PHH imaging

Axiom

AI model

ROC-AUC--0.85
Sensitivity / True Positive Rate41%71%
Specificity / True Negative Rate86%91%
Cost$3,000–$5,000$100–$450
SOURCE "Evaluation of the use of imaging parameters for the detection of compound-induced hepatotoxicity." S Schadt et al, Toxicol In Vitro, 2015

Inside our model: how it works

Axiom uses molecular structure, properties, and cell biology to predict clinical toxicity outcomes.

Proprietary data combined with the latest ML/AI methods

We combine the world's largest human liver dataset with the latest AI/ML methods to train highly accurate models.

Step 1

We’ve created the world’s largest primary human liver dataset

  • 115,000+ small molecules exposed to primary human liver cells.

  • High content imaging of key cell organelles paired with biochemical assays.

  • Learn more about Axiom's training set.

Step 2

We quantify biology with unprecedented precision

  • 10+ models of unique biology (confluency, ER-stress, mitochondrial toxicity, and more).

  • Bile canalicular networks, efflux/uptake transporters, steatosis coming soon.

  • Full, 8-point dose-response curves.

Step 3

Our models learn how chemistry affects biology

  • Molecules induce diverse cellular phenotypes.

  • Axiom's models learn which molecules induce which toxic cell phenotypes.

  • For a new molecule, our models identify its cellular phenotype and clinical toxicity risk, and how these vary with dose.

Assess toxicity in a modern web application

Reason about clinical liver injury risk with powerful data visualization and predictions.

How to use our model

Predict, interpret, optimize

Key Features

An easy end-to-end workflow.

Pricing

Axiom's Model vs CRO Experiments

Axiom

Axiom

AI Model

$100–$450
per molecule

Other

Physical experiments

2D PHH, 3D Spheroid, Organ-on-a-chip, Animals, etc

$5k–$20k
per molecule