Within the framework of the Operational Program ERDF 2021-2027, ACTION 1.1.2 “Support for collaborative research projects of research laboratories and universities with enterprises” of the Emilia-Romagna Region, Call for Laboratories 2023-2024 for strategic industrial research projects addressing the priority areas of S3 2023-2024, IRET Foundation is the lead partner of two projects: AI4ChemoBrain and N2ERT.
AI4ChemoBrain – A machine learning/artificial intelligence model for personalized medicine: development of a predictive marker for cognitive complications during chemotherapy based on subject-derived phenotype and omics data
➡️ Start date: 06/02/2024
➡️ End date: 06/08/2026
⏱️ Total duration: 30 months
The term ‘chemobrain’ denotes a cognitive impairment, which arises in the majority of patients during and after chemotherapy. Although still poorly characterized, systemic inflammation and intestinal dysbiosis are regarded as possible causes. The project aims to generate a prototype demonstrator predictive of the onset of chemobrain, for clinical use, based on a machine learning/artificial intelligence model trained and validated with preclinical data sets. The ML/AI model will be trained using historical descriptive data of cognitive decline, the quality of which can be verified by fully tracking their life cycle, optimising their value. The ML/AI model will then be validated using data sets collected from chemobrain mouse models generated ad hoc for the project. The ‘data-driven’ approach exploited by the ML/AI model will also be validated by parallel analysis of all data sets with conventional statistics (‘hypothesis-oriented approach’). The ultimate goal is the translation of the preclinical model to the clinic, with the release of a ML/AI demonstrator predictive of the chemobrain condition in cancer patients in order to promote adjuvant therapies that can also be exploited in the perspective of ‘personalised medicine’.
Objectives and expected results
A prototype predictive model of cognitive decline associated to oncological therapies, and based on supervised and unsupervised ML/AI techniques will be developed.
The project consists of three main activities:
-
- ML/AI model training: historical preclinical cognitive and omics data will be used according to supervised learning techniques;
- ML/AI model testing: the model obtained in activity 1., will be tested with a data set derived from a preclinical chemobrain model, in order to assess the predictive ability of the chemobrain (unsupervised learning techniques) in comparison to conventional and correlation statistics;
- ML/AI model validation by:
-
-
- Data set derived from a second preclinical model of chemobrain in subjects with cognitive decline;
- Data set derived from an external cohort.
-
The combination of these data sets will make it possible to study the contribution of each descriptor to the model’s predictive ability. Data augmentation, transfer learning and fine-tuning techniques may be used to overcome possible problems related to the availability of large data sets within the chemobrain to improve the predictive capabilities of the ML/AI model.
The partnership brings together laboratories of the High Technology Network (IRET Foundation, CIRI SdV – University of Bologna, CIRI ICT – University of Bologna, Biopharmanet TEC – University of Parma, Laboratorio per le Tecnologie delle Terapie avanzate LTTA – University of Ferrara) and companies (Alfasigma Spa, Wellmicro Srl), leaders in their respective fields, which are a good guarantee of achieving the project objectives.
The AI4ChemoBrain project is co-funded by the ERDF RP 2021-2027, ACTION 1.1.2 of the Emilia-Romagna Region.
👉🏻 To keep up with project updates you can visit https://ai4chemobrain.it/en/home/.
