DIOPTRA

Early Dynamic Screening for Colorectal Cancer via Novel Protein Biomarkers Reflecting Biological Initiation Mechanisms

Consortium28 partners · 15 countries

RoleTechnical Partner — AI Modelling

ProgrammeHORIZON EUROPE

Grant reference101096649

Our role

Technical Partner — AI Modelling

We lead the project's AI modelling, identifying the protein biomarkers with the highest discriminative power for colorectal cancer screening. The goal is a small, cost-effective set of proteins that also accounts for risk factors and comorbidities, and we deliver the interim and final DIOPTRA AI models.

Impact and outcomes

Targeting more than 85% specificity for AI-assisted colorectal cancer screening
Targeting more than 1,200 participants recruited across 8 clinical sites, 40–50% of them women
Targeting an increase of at least 15% in screening participation within 5 years of project completion
Discriminative protein biomarker subset planned, backed by network analysis of its biological role

The challenge

Colorectal cancer is the second leading cause of cancer deaths, with 1.9 million new cases in 2020 and a projected 3.2 million a year by 2040. Five-year survival reaches 90% at stage I but falls below 15% at advanced stages. Yet current screening is either invasive, like colonoscopy, or less reliable, which keeps participation low.

Our approach

DIOPTRA screens a pool of 3,000 candidate proteins in blood samples to find a discriminative subset for early detection. It combines them with risk-factor data and validates the protocol in pilot studies at 8 clinical sites. Our AI models identify the most informative biomarkers and risk factors in an explainable way, moving beyond black-box predictions.

Website →