Scientific Basis:
The WHO considers autism as a priority due to its high-frequency, early onset, and life-long impact on quality of life. Discovering mechanisms that predict which autistic person develops mental health problems (prognostic biomarker), and which treatment may benefit individuals (predictive biomarker) requires candidate biomarkers derived using the most robust predictive methods. This PhD leverages data from the multi-centre Longitudinal European Autism Project, including clinical, cognitive, neuroimaging, environmental and genetic data from ~700 individuals (430 autistic) aged 6-35 years.
Computational modelling distils multivariate neuroimaging data into parsimonious parameters. One approach is to summarise whole-brain connectomes from magnetic resonance imaging (MRI) or electro-encephalography (EEG) using graph theory. Another emerging approach is to quantify and model spatial dependencies in neuroimaging data. These complex methods provide mechanistic understanding of anatomy and connectivity in health and disease. However, their added prognostic value over simpler imaging or cognitive measures is unclear.
Aims & Yearly Objectives:
Compare predictive utility of graph theory and spatial dependencies to regional anatomy, activity and cognition for predicting the disease course of autism spectrum disorders. Identify individual connectivity and spatial dependency profiles and test their relation to concurrent or future development of mental health problems (controlling for environmental risk and genetics).
Year 1: Constructing connectomes and quantifying spatial dependencies from the EU-AIMS LEAP project.
Year 2: Application of machine learning algorithms to define candidate biomarkers.
Year 3/4: Test added value of connectivity modeling for prognostics in longitudinal dataset.
Techniques:
This project bridges cognitive neuroscience, computational modelling and machine learning, and embeds the student within a vibrant research community with leading experts on autism and neuroimaging.
