PhD Studentship: Integrative Analysis of ADHD Using Graph Kernels on UK Biobank Multimodal Data

Introduction

Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder marked by inattention, hyperactivity, and impulsivity. Globally, it affects about 5% of children, with 30–50% continuing to experience symptoms into adulthood, leading to challenges in education, employment, and relationships. In the UK, the National Institute for Health and Care Excellence (NICE) estimates that ADHD affects around 3–4% of adults. The Centre for Mental Health has estimated that each case of ADHD imposes a lifetime cost of approximately £100,000 on society. Implementing effective preventive treatments could reduce these expenses, leading to improved individual outcomes and significant economic savings. Despite its widespread impact, the precise etiology of ADHD remains elusive, necessitating comprehensive research to unravel its complex biological underpinnings.

The Imperative of Data Integration

ADHD's etiology is multifaceted, involving genetic, environmental, and neurobiological factors. Analyzing each data modality in isolation may overlook critical interactions and lead to inconclusive findings. The UK Biobank provides a comprehensive biomedical dataset from half a million participants, invaluable for ADHD research. It spans multiple modalities, including demographic information, cognitive assessments, genetic data (SNPs), proteomics, and neuroimaging. This dataset supports holistic analyses of ADHD's multifactorial etiology. Integrating these diverse datasets is essential to enhance signal detection amidst the inherent noise of biological data. A unified analytical approach can uncover complex relationships and provide a more comprehensive understanding of ADHD's pathophysiology.

Model integration:

Graph Kernels and Graph Neural Networks as a Flexible Approach to Multimodal Data Integration Graph-based methods effectively model complex biological systems by representing entities as nodes and their interactions as edges. Graph kernels enable comparisons by capturing structural similarities and can be tailored to specific data modalities: modeling SNP interactions in genetics, protein networks in proteomics, and brain connectivity in neuroimaging. Integrating these modality-specific kernels Engineering and Physical Sciences Research Council Doctoral Landscape Award creates a comprehensive model to uncover ADHD biomarkers and therapeutic targets. Furthermore, feature descriptors derived from each graph kernel can be integrated using deep graph neural network techniques, enhancing the model's capacity to learn complex, non-linear relationships across diverse data types. Research Objectives

This PhD project aims to:

1. Develop Modality-Specific Graph Kernels: Design and implement graph kernels tailored to each data modality within the UK Biobank dataset.
2. Integrate Multimodal Data: Combine these kernels to create an integrative model that captures the complex interactions underlying ADHD.
3. Identify Biomarkers: Utilize the integrative model to discover biomarkers predictive of ADHD diagnosis and progression.
4. Enhance Understanding of ADHD Pathophysiology: Provide insights into the biological mechanisms contributing to ADHD, informing future research and therapeutic strategies.

Expected Outcomes

The proposed project aims to develop an analytical framework that integrates diverse biological data to elucidate the complex etiology of ADHD. Anticipated outcomes, suitable for dissemination through scholarly publications, include: 1. Development of customized graph kernels for different data modalities, highlighting their utility in ADHD research. 2. Integration of these kernels into a unified model using graph neural networks (GNNs) to identify biomarkers and understand ADHD's pathophysiology. 3. Evaluation of the integrated model's predictive accuracy for ADHD diagnosis and progression using UK Biobank data, compared to single-modality methods.

Advert information