Identify multi-omics biomarkers with Abzu's explainable AI

You can find the simplest, and otherwise hidden, cause or causes for heterogeneous and complex disease mechanisms with Abzu's explainable AI.

Public healthOncologyWomen’s healthMulti-omics
Learn more

We identified two gene expressions as primary mortality indicators on a wide data set of only 705 women

Biomarker discovery: too slow for too many

Early prognosis directly impacts outcome. We wanted to find biomarkers that accurately predicted mortality outcomes and better understand this heterogeneous and complex disease.

Explainability will accelerate your understanding

Our predictions were not only more accurate than traditional machine learning methods, but our simple models also revealed specific genetic indicators for a poor outcome.

Join us on the next scientific revolution

We are confirming our findings in the latest research alongside oncology specialists. If you’re conducting breast cancer research then we want to collaborate with you.

Your next discovery has the potential to change the current state


13% of women will develop breast cancer during their lifetimes. [1]


Incidence rates of breast cancer have increased by 0.5% every year. [1]


Breast cancer is the second leading cause of cancer death in women. [1]


Early detection directly decreases the proportion of poor outcomes. [1]


So this is real new scientific knowledge, right? This is not about predicting breast cancer. This is about explaining what it is, actually, that causes breast cancer to be lethal in some women, whereas, fortunately, most women get through it with their life and health intact.

Casper Wilstrup, CEO @ Abzu

Simplify your biomarker strategy without sacrificing accuracy

Out of almost 2.000 features, a simple and explainable model identified just two specific gene expressions (the proteins APOB and MYOC) as causal pathways for poor outcomes.

Black-box predictions aren't science

To understand the basis of disease and identify novel points of intervention, scientists need to employ an AI that generates high-performing and explainable models. Abzu delivers explanations that scientists and researchers – real humans – can actually understand.

The data set

We used data collected and processed by The Cancer Genome Atlas (TCGA)[2], a landmark cancer genomics program. This public data set is available through the National Cancer Institute's Genomic Data Commons (GDC)[3]. The multi-omics data set is composed of only 705 samples with 1.936 features consisting of genetic copy number variations, somatic mutations, gene expressions, and protein expressions.

Abzu's "Smart Biomarkers"

Abzu's data science and full packages include our explainable AI and flexible data science, data generation, and scientific support.

Data quantification
Data preparation
Abzu’s QLattice®
AI modelling

The team

This research was completed in-house at Abzu by Sam Demharter, PhD in Systems Biology, and Valdemar Stentoft-Hansen, MSc in Economics.


Abzu’s explainable AI unravels disease mechanisms

We only understand a fraction of the human body. Let’s know more.

[1] American Cancer Society. Key Statistics for Breast Cancer. 2022. https://www.cancer.org/cancer/breast-cancer/about/how-common-is-breast-cancer.html

[2] The Cancer Genome Atlas Program. 2022. https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga

[3] National Cancer Institute Genomic Data Commons. 2022. https://gdc.cancer.gov/