The list of genes contributing to increased risk of developing Alzheimer’s Disease (AD) grows every year. To date, large case-control Genome-Wide Association Studies have found 98 genetic loci associated with late-onset Alzheimer's Disease (LOAD). However, apart from a small number of genes related to familial cases, it remains largely unknown the direct relationships between genes amyloid-β and tau misfolding protein accumulation (hallmark of AD), and neurodegeneration of the human brain. Even more striking, current knowledge of mutations potentially conferring genetic resilience is sparse. A better characterization of each genetic risk loci and identification of protective genetic variants could lead to development of effective therapeutical targets, actually unavailable for AD. The goal of this project is to characterize associations between the 98 LOAD genetic risk loci, in vivo tau and amyloid-ß PET, and MRI neurodegeneration – including an examination of potential genetic variants conferring disease resilience. Hypotheses: genetic loci will show distinct spatial patterns of tau and amyloid-ß spreading and neurodegeneration, helping us to better understand AD biological heterogeneity. I expect to find different subtypes of AD with different molecular mechanism and different therapeutical targets for each subtype. I will use PET imaging of ~3,000 participants from the Harvard Aging Brain Study, and the A4 and ADNI databases. ~35,000 participants from the UK Biobank with genetic and functional and structural MRI will also be used to study the impact of genetic loci on neurodegeneration. Importantly, as a first analysis, we will search for the effects of each risk loci and protective variants conferring resilience; secondary analyses will also explore the contributions of sex and ethnic minority factors

Our overall objectives are to accelerate the diagnosis, and enable personalised management, of inherited metabolic diseases (IMDs). Established academic technology for statistical genomic analysis, deep learning-based prediction of protein structure, and whole-body metabolic network modelling shall be applied to generate personalised computational models, given patient-derived genomic, transcriptomic, proteomic and metabolomic data. To train diagnostic models, a comprehensive clinical team will recruit 1,945 diagnosed patients with a wide variety of IMDs, then validate the clinical utility of personalised computational models on a set of 685 undiagnosed patients. An enhanced human metabolic network reconstruction, especially for lipid metabolism, reaction kinetics and inherited metabolic disease pathways, will increase the predictive capacity of cellular and whole-body metabolic network models. As an exemplar for other IMDs, personalised computational modelling will be used to identify compensatory and aggravating mechanisms that associate with clinical severity in Gaucher disease. The predictive capacity of personalised models will be validated by comparison with additional empirical investigations of protein structure and function as well as metabolomics, tracer-based metabolomics and proteomics of patient-derived in vitro disease models. To maximise the potential for impact, personalised modelling software will be developed to be generally applicable to a broad variety of IMDs, and implemented in a way that is both accessible to clinicians and admissible to regulatory authorities. Sustainability will be promoted by development of a roadmap for a European foundation to aid personalised diagnosis and management of IMDs, informed by broad stakeholder consultation. This is a unique opportunity to realise the potential of personalised computational modelling for a broad set of rare diseases, which is a field where European collaboration is an essential for progress.

AISym4Med aims at developing a platform that will provide healthcare data engineers, practitioners, and researchers access to a trustworthy dataset system augmented with controlled data synthesis for experimentation and modeling purposes. This platform will address data privacy and security by combining new anonymization techniques, attribute-based privacy measures, and trustworthy tracking systems. Moreover, data quality controlling measures, such as unbiased data and respect to ethical norms, context-aware search, and human-centered design for validation purposes will also be implemented to guarantee the representativeness of the synthetic data generated. Indeed, an augmentation module will be responsible for exploring and developing further the techniques of creating synthetic data, also dynamically on demand for specific use cases. Furthermore, this platform will exploit federated technologies for reproducing un-indentifiable data from closed borders, promoting the indirect assessment of a broader number of databases, while respecting the privacy, security, and GDPR-compliant guidelines. The proposed framework will support the development of innovative unbiased AI-based and distributed tools, technologies, and digital solutions for the benefit of researchers, patients, and providers of health services, while maintaining a high level of data privacy and ethical usage. AISym4Med will help in the creation of more robust machine learning (ML) algorithms for real-world readiness, while considering the most effective computation configuration. Furthermore, a machine-learning meta-engine will provide information on the quality of the generalized model by analyzing its limits and breaking points, contributing to the creation of a more robust system by supplying on-demand real and/or synthetic data. This platform will be validated against local, national, and cross-border use-cases for both data engineers, ML developers, and aid for clinicians’ operations.

Chemistry and Biology are governed by molecules and how they interact. Crucially, what glues a molecule together are chemical bonds, made from atoms pairing all their electrons. Although preferred, this is not the only option: in the comparatively rare cases where a molecule presents unpaired electrons, it acquires a fascinating new status that transforms its chemical and biological properties, best described by the acutely apt name of radical. Despite the extraordinary toolset found in radical-bearing molecules, the rather demanding methods to radical formation currently available mean that only very specific molecular architectures can withstand them, inadvertently limiting the scope of their applicability. The aim of this ERC project is to show that reversible diradical formation upon deprotonation is prevalent, and yet unexplored, in general donor-acceptor organic molecules and use this new knowledge to develop novel design criteria in light-emitting molecules and drug discovery. To achieve this unique aim, I will exploit a widespread structural pattern in a novel way, enabling a molecule to reversibly convert its charge and spin and become a diradical. I will first characterise how different molecular constituents (un)favour diradical formation on isolated molecules. I will then establish, for the first time, the role that diradicals play in defining the function of the numerous bioactive molecules sharing the proposed structural pattern. By exposing the overlooked diradical character in general families of deprotonated organic molecules, I will deliver transformative mechanistic understanding on i) the photo physical properties of fluorescent proteins and ii) the reactivity of small molecule drugs, particularly a new class of covalent inhibitors. The field of organic radicals sits at a critical crossroads between Chemistry and Biology, and as such, taking it a step forward has the potential to cross-pollinate research fields and reshape research frontiers.