Deep Learning Pipeline of Multi-Modal Biomarkers for Alzheimer’s Disease Early Detection and Progression

This project aims to develop and validate an AI-driven composite info-biomarker panel for Alzheimer’s disease (AD), combining extracellular vesicle (EVs)-derived proteomic signatures, demographic data, and genetic risk factors. By applying EVs biology as a window into neuronal pathology, and advanced machine learning, the ‘composite info-biomarker’ panel are expected to detect early AD’s changes and monitor progression. The work will deliver point-of-care–compatible EVs isolation platform and a deployable decision-support pipeline for clinical and trial settings.
The PhD student will be based at the Institute of Pharmaceutical Science. The student will have access to state-of-the-art equipment and tailored training (e.g., 3D printing techniques, nanoparticle characterisation, data analysis, artificial intelligence and bioinformatics) supported by mentorship schemes. The student will be supervised by three experts with complementary skills: Dr. Mohamed A Alhnan, with over 20 years’ experience in 3D printing, drug delivery systems, and pharmaceutical technology and Prof. Dag Aarsland, an experienced clinician-scientist specialising in old age psychiatry and dementia research.
The purpose of this studentship is to develop and train interpretable machine learning models that fuse exosomal biocorona proteomic signatures with clinical and genetic data, enabling early AD detection and disease progression monitoring.
Year 1 will focus on establishing necessary infrastructure and essential skills: setting up laboratory workflows for nanoparticles isolation and characterisation, obtaining ethical approvals and protocols for sample handling, and building basic data-processing pipelines. The PhD student will conduct initial pilot experiments to verify methods and learn core data analysis techniques, while attending relevant training workshops and defining a data management strategy.
Year 2 will involve generating and processing a larger dataset, preparing feature matrices, and exploring preliminary analytical models. The PhD student will apply representation and classification approaches to integrate biomarker and patient data, interpret early results to refine candidate markers, and plan a validation study.
Year 3 will emphasise model validation and generalisation: testing approaches on independent or collaborative datasets, adapting models as needed, and engaging with stakeholders or advisory groups to pilot the workflow, collect feedback, and adjust methods accordingly.
Year 4 will address longer-term sustainability: implementing processes for ongoing monitoring and model updating if any new data updates, and finalising analyses. The PhD student will consolidate results and outline future directions for translational application or broader studies.