BioMarkML: Machine Learning-Driven Bio-Sensing for Early Diagnosis of Alzheimer's Disease

Alzheimer's Disease (AD) represents approximately 60% of all dementias and affects over 33 million people globally costing healthcare systems 1% of global GDP to diagnose, treat and care for patients. In the UK over 0.6 million people suffer from AD costing over £16b/y. Therapeutic strategies depend on the diagnosis of AD as early as possible to prevent cognitive deficits before they occur.

In this PhD project we will exploit machine learning (ML) techniques to enable classification of case vs control data sets for protein AD biomarkers using multiplexed graphene sensors that have already been demonstrated to have ultra-high sensitivity and specificity. The ML stage will involve; (i) Feature Selection: pre-processing bio-signals (conductance & impedance) for analysis, employing techniques to iteratively remove the least important features and prevent model overfitting, and utilizing ML models to assess and rank feature importance; (ii) Pattern Discovery: through clustering approaches to understand the underlying patterns of bio-signals and to inform the classification process; (iii) Classification: to differentiate Alzheimer's patients and healthy controls; (iv) Validation: to rigorously evaluate the classification model's performance and ensure its reliability and generalizability, and (v) Optimisation: tuning hyper-parameters or refining the data pre-processing to optimize the model's performance.

The pioneering nature of this research will enable integration of ML algorithms with graphene sensors to realise a Point-of-Care (PoC) System. The successful candidate will join a multidisciplinary team spanning AI/ML, biochemistry/biophysics, electronics and clean room-based fabrication of sensors, including access to facilities and expertise from our industrial partner Zimmer & Peacock Ltd.