Federated artificial intelligence for privacy-preserving international stratification of colorectal cancer patients

Project Information

Microb-AI-ome

Grant agreement ID: 101079777

DOI

10.3030/101079777

EC signature date 16 March 2023

Start date 1 April 2023

End date 31 March 2028

Funded under

• Health

Total cost

€ 5 996 700,25

EU contribution

€ 5 996 700,25

Coordinated by

UNIVERSITY OF HAMBURG
Germany

Periodic Reporting for period 1 - Microb-AI-ome (Federated artificial intelligence for privacy-preserving international stratification of colorectal cancer patients)

Reporting period: 2023-04-01 to 2024-09-30

Summary of the context and overall objectives of the project

In the EU, 1 in 35 women and 1 in 23 men will be diagnosed with colorectal cancer (CRC) in their life span (ca. 340,000 cases and 156,000 deaths in 2020) causing an annual economic burden of ca. 20 billion EUR. Identifying CRC early enables better treatment options. Screening usually entails a quantitative faecal immunological test (FIT) to predict the need of colonoscopy for the detection of colorectal lesions, an expensive and invasive procedure. We aim to predict this need with specificity increased by >20 percentage points by using metagenomic microbiomes. We hypothesise that computational microbiome profiles extracted using artificial intelligence (Al) technology will also allow for optimised personal therapy stratification. However, clinicians do not have access to broad microbiome datasets. With Microb-AI-ome, we will develop a novel kind of computational stratification technology to enable microbiome-enhanced precision medicine of CRC. Metagenomic microbiome data that has been generated to date is distributed over many national registries, and privacy regulations are hindering its effective integration. With Microb-AI-ome, we will overcome this barrier by establishing the first privacy-preserving federated big data network in CRC research. We will integrate isolated, national databases into one international federated database network - rather than a cloud - covering metagenomes for over 5,000 individuals screened for CRC, and an expected total of 100,000 by 2026. Microb-AI-ome ensures that no sensitive patient data will leave the safe harbours of the local databases while still allowing for the classification of clinical CRC phenotypes, which we will demonstrate in clinical practice allowing regulatory bodies to adopt evidence-based guidelines. Our consortium combines expertise in CRC and its treatment, microbiomics, artificial intelligence, software development, and privacy protection to close the gap between privacy and big data in international medical research.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

In the first project period, we implented a prototype for the local databases that are to be joined in the federated database network Microb-AI-Net. Each of these local nodes, or Microb-AI-Clients, will serve as a data node within the federated database network and allow the federated training of AI models, the so-called Microb-AIs. To ensure data is added to the Microb-AI-Clients in a harmonized manner, data standards were defined and implemented by the consortium. In parallel, all relevant ethical and regulatory documentation for clinical data collected, as well as a data management plan, were assembled, ethics approvals were gained, a study initiation package was published and patient recruitment commenced in both France and Ireland. An ethical and human rights impact assessment framework was created in order to help this process and provide legal and ethical guidance for the project. While data collection is ongoing, the work on the training of Microb-AIs for the prediction of CRC based on microbiome-sequencing data has started with the help of publicly available datasets. To this end, a range of microbiome analysis tools has been evaluated and an analysis pipeline has been agreed upon.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Microb-AI-ome adresses all three of the expected outcomes as stated in the work programme:

Clinical researchers use effective health data integration solutions for the classification of the clinical phenotypes - Microb-AI-Net will provide an infrastructure to CRC clinicians and researchers that, for the first time, allows the integrated analysis of international microbiome and healthcare data for the training of CRC stratification models.

Researchers and/or health care professionals use robust and validated data-driven computational tools to successfully stratify patients - The trained Microb-AIs for prediction of a colonoscopy based on microbiome data will be validated for their robustness and stratification performance as part of the project and made available through the CRC Stratifier software. Once validated, these models can be integrated into medical guidelines, where appropriate.

Regulatory bodies approve computer-aided patient stratification strategies to enable personalised diagnosis and/or personalised therapy strategies. Health care professionals adopt evidence-based guidelines for stratification-based patient management superior to the standard-of-care. To prepare for the clinical deployment of Microb-AI-ome beyond the project runtime, we will develop protocols that comply with both EU and US regulatory standards, particularly in the final phase of clinical deployment. We will seek in-house accreditation for our protocols according to ISO 15189 and Clinical laboratory ISO 13485 standards for in vitro diagnostics. This will ensure that we meet the relevant CLIA and EU lab accreditation requirements, EU/CE mark standards, and FDA requirements. During the final clinical validation phase, we will implement a Quality Management System (QMS) and data handling procedures to comply with ISO 13485, ISO 15189, and 21 CFR 820, as well as Regulation (EU) 2017/746. We will also ensure compliance with evolving medical diagnostic regulations and with GDPR requirements.