D2.17: Neurodegenerative use cases: Intermediate results

Le résumé fourni par la source

This deliverable reports the intermediate results of the HEREDITARY neurodegenerative use cases, with a focus on Use Case 1, dedicated to neurodegenerative disease phenotyping and prognosis evaluation, and Use Case 2, dedicated to next-generation diagnosis and treatment-response modelling. ALS is used as the primary model disease because of its clinical, genetic, neuroimaging and biological heterogeneity, making it a suitable demonstrator for multimodal, ontology-enabled and privacy-preserving data integration. D2.17 documents the progression from use-case design to data readiness, biological interpretation and a roadmap for federated multimodal analytics. The deliverable builds on FAIRification activities described in D3.6, including HealthDCAT-AP metadata records for ALS resources, and on the Beacon v2 pilot reported in D3.11, which demonstrated privacy-preserving genomic discovery in the UNITO ALS whole-genome sequencing cohort. Initial semantic harmonization between Piemonte and Valle d’Aosta Register for Amyotrophic Lateral Sclerosis (PARALS), AnswerALS, Beacon v2 and the HEREDITARY ontology, HERO, confirmed the need for ontology-mediated integration across heterogeneous ALS datasets. The deliverable also consolidates recent biological findings from HEREDITARY partners, including evidence of sex-specific genetic liability in ALS, gene-related FDG-PET metabolic signatures and sex-related differences in brain metabolism and cognitive reserve. These results support a multidimensional model of ALS heterogeneity in which genetic liability, sex-modified penetrance, brain network vulnerability, cognitive reserve and clinical progression interact. Based on this framework, D2.17 proposes that ALS stratification should move beyond direct clustering of raw variables towards latent multimodal representation learning. The proposed strategy includes modality-specific embeddings, ontology-guided alignment through HERO, Linked Independent Component Analysis (LICA), supervised multimodal decomposition, Event-Based Models (EBM) and Subtype and Stage Inference (SuStaIn). Together, these approaches define the methodological basis for ontology-enabled ALS patient stratification and future federated multicentric implementation. At this intermediate stage, D2.17 verifies Milestone 9 (Neurodegenerative diseases phenotyping and prognosis evaluation) and sets the ground for the subsequent work to be consolidated in D2.18. It shows that HEREDITARY has progressed from use-case definition towards operational data discoverability, genomic Beaconization, semantic mapping, biologically informed multimodal modelling and a concrete roadmap for federated ALS clustering.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.