Characterising the nature of mental health trajectories across adolescent development through the integration of genomic, biomarker, neuroimaging and

Poorer mental health in young people contributes to 45% of all years lost because of disability. Mental health disorders will often onset during this age, resulting in concurrent and long-lasting inequalities with implications for the patient, family and society. Thus, there is an essential need to identify early interventions and preventions that combat both the disorder and downstream consequences.
However, not all current interventions and treatments work. One key challenge is that mental health disorders during this period change considerably, and little is known about how specific risk factors underpin the timing, course and severity of poorer youth mental health trajectories. Greater understanding of the mechanisms underpinning these trajectories (such as the interplay between genetic risk, epigenetic changes, biomarkers brain changes, neurocognitive impairments and environmental risk factors) could greatly enhance our ability to deliver successful interventions and preventions to the right people at the right time.
Longitudinal studies are an important tool for examining how mental health changes over time and for whom. This project will use newly available large longitudinal datasets to examine how a combination of genetic risk, epigenetic changes, environmental risk factors, brain imaging and neurocognitive data, and biomarkers predicts different mental health trajectories across childhood, adolescence and early adulthood. Data included are: the Adolescent Brain Cognitive Development study (ABCD; n=~11,900; age 9-14 years), the Avon Longitudinal Study of Parents and Children (ALSPAC; n~15,600; age 0-30 years) and the Teds Twins study (TEDS; n=~10,000; age 0-25).
This project will apply machine learning approaches to model heterogenous mental health trajectories of mental health and identify variables that have the most influence on such prediction models. Variables will include genomic (e.g., polygenic risk scores), neuroimaging (e.g., functional connectivity), inflammatory (e.g., blood biomarkers) and environmental data (e.g, traumatic events) that has been collected at a series of time points within these longitudinal datasets. Machine learning provides a powerful computational approach that can combine features that display multicollinearity and classify groups based on the data distribution and predictive variables. Building predictive models of mental health outcomes will catalyse the field of precision psychiatry by stratifying patient groups. As a result, the right treatment can be distilled for the right person at the right time - the crux of precision medicine.