Digitalized AI tool for safety prediction of ATMPs (digiSafe-ATMPs)

Purpose and goal:

Advanced Therapy Medicinal Products (ATMPs) is a rapidly emerging field in medicine with great potential. This project addresses the strong need for robust and scalable safety prediction methods for industrial ATMP manufacturing and delivery, that are required for marketing authorisation of ATMPs with great benefits for patients. Today, scalable, standardised, and validated methods are lacking for:

- assessment of patient safety, such as scoring the risks for graft rejections - safe and efficient delivery of ATMPs to specific injured or diseased tissue Expected results and effects:

New safety biomarkers to identify abnormalities at early stages of the manufacturing process of ATMPs will be developed. This will promote advanced digitalisation in the ATMP sector, thereby improving weaknesses in today´s standard methods for predicting quality problems that can potentially lead to large losses of resources and a reduced quality of life for patients.

Approach and implementation:

AI models will be trained with omics datasets generated from various stages of the fabrication of individualised blood vessels transplanted into laboratory animals. Siamese Neural Networks-models will be developed to calculate a similarity score to native blood vessels to assess the safety of ATMP and improve safety assessment tests so that successful transplants can be predicted.

Furthermore, Variational Autoencoders will be developed and trained on integrated omics data to identify new safety biomarkers.