The ETHEREAL project includes four main scientific activities:
1. The identification of environmental factors that “causally” could instigate in an autistic adolescent a transition of an already vulnerable “mental health” to “mental illness” (Model 1);
2. A quantitative modelling of the Epigenetic-Genetic/ metabolomic-Mental health (EGM) process and the design of control strategies (Model 2); it is the first quantitative causal model characterizing the EGM process chain and represents the missed gap for an effective and personalized management of mental health in adolescents.
3. The Integration of these models in a mobile app (the Personal Digital Nurse, PDN), that empowers the adolescents to self-control the influence of the factors present in his/her Personal Natural Environment (PNE);
4. The design of a new and personalized model of service delivery based on Implementation Science and capable to link the adolescent with the community’s care services (Model 3) and to provide a set of neurobiologically informed criteria for the selection of the most appropriate EBIs.
An observational study (ENIGMA study) in Ireland (Cork), Romania (Bucharest) and United Kingdom (Southampton) allowed to generate heterogeneous sets of data that were used in the modelling activity. To capture nonlinear causal relations between the environmental factors, metabolites and anxiety and depression of the adolescents we applied a novel technique, the Synergistic-Unique-Redundant Decomposition (SURD) model of causality, decomposing the causal relations between the variables in three parts (unique, synergetic and redundant causalities) and representing them as knowledge graphs at individual level and at group level. The most relevant PNE factors in the EGM process chain were identified and how the process could be controlled by influencing the modifiable PNE factors.
As part of the activity, a novel AI-enabled classifier was designed that predicts the level of depression and anxiety as a traffic light system (low, moderate and high level). It use dendritic artificial neural networks operating in the frequency domain, not in amplitude domain that represents the cornerstone of the current generation AI.
The care delivery system designed by ETHEREAL (Model 3) is based on a 3-tier community and clinic services that involve key stakeholders and represents a holistic approach and an example of community-based care model, that aims at filling the gap of the currently heavy fragmentation of health- and social care services.
At the core of this system there is a mobile app, the Personal Digital Nurse (PDN) with an AI engine (a Graph Neural Network (GNN)) where the EGM model is translated along with the functionalities required for holistically supporting the autistic adolescent in predicting and avoiding potential trajectories towards depression and anxiety and for facilitating the access to the most effective personalised EBIs, based on the understanding of the cognitive strengths and weaknesses of the adolescent that will inform the choice of the EBIs. In the final part of the project, a multidimensional assessment including usability tests and involving a large and international group of key stakeholders and experts allowed to get useful inputs for further development and refinement.
