Digital avatar in bipolar disorder, a precision medicine approach for clinical decision support

Digital avatar in bipolar disorder, a precision medicine approach for clinical decision supportThe main aim is to develop a Digital Avatar solution to predict disease onset and outcomes in bipolar disorder (BD), building on large samples, novel machine learning tools, and clinical validation. For this reporting year, progress was made on harmonizing datasets, method development, and integration with clinical and biobank datasets.The Digital Avatar project for bipolar disorder (BD) is focused on developing a precision medicine tool that integrates multimodal big data—such as genetic, brain MRI, cognitive biomarkers, and clinical data—to improve the prediction of disease onset, treatment response, and management. This tool, the "Digital Avatar," aims to create a personalized model for each patient, enabling more accurate treatment strategies tailored to individual needs. This past year, the project has made important progress, primarily in method development and sample and biobank expansion. We have focused on method development, including building our predictive models using machine learning (ML) and Bayesian statistical techniques to increase their accuracy. This includes ongoing work on integrating large datasets from the EU H2020 projects R-LINK and RealMent, which provide crucial insights into lithium treatment response, disease onset, and comorbidities such as cardiovascular disease. Additionally, we have prioritized the recruitment of new individuals to the biobank, enhancing the size of our dataset, as large samples are needed for prediction. Efforts have also been directed towards harmonizing the samples to ensure compatibility and consistency across data sources, crucial for accurate model development. Moreover, we are working on expanding the model to predict not only BD onset but also comorbidities such as cardiovascular disease (CVD), which is prevalent among BD patients and contributes to the high mortality rate. By incorporating data on CVD risk factors and outcomes, we are developing a comprehensive tool for BD management. This includes building a CVD tool, including a polygenic hazard score, which can be applied in patients with BD, to stratify according to risk. The next steps will involve further model refinement and the continued expansion of our sample and biobank. These efforts will lay the foundation for a clinically relevant tool that can guide individualized treatment decisions for BD. Our long-term goal is to transform BD care through precision psychiatry, improving patient outcomes, quality of life, and reducing healthcare burdens.