Dr Mamunur Rashid

Supervisor Details

Contact Details

Dr Mamunur RashidLink opens in a new window

Department of Cancer and Genomic Sciences, University of BirminghamLink opens in a new window

Scientific Background

Research Interests

My current research is deeply anchored in digital health equity and explainable AI. By actively exploring the diversity dimensions of multimodal healthcare data, I employ knowledge-driven approaches to augment data gaps and mitigate inherent biases. This critical work ensures that the resulting predictive models are fair, equitable, and transparent, enabling their safe clinical implementation to benefit all patient demographics equally.

Scientific Inspiration

Professor John Quackenbush serves as a profound scientific inspiration due to his pioneering leadership in bridging computational biology, biostatistics, and clinical medicine to tackle complex human diseases. His career exemplifies a seamless interdisciplinary trajectory - spanning from foundational genomics and molecular biology to large-scale data science and clinical translation.

Research Groups

Birmingham AI & Digital Health and Care Group (Bright)Link opens in a new window

Supervision Style

In three words or phrases: Lead by example, collaborative, goal-driven.

Provision of Training

I will train and mentor you as required, ensuring you also learn from other experienced members of my group. My supervision goes beyond standard academic guidance; I actively teach students Agile Project Management principles and train them to build robust AI data analysis frameworks, equipping them to transition smoothly into competitive industrial roles or further research.

Progression Monitoring and Management

We will embark on this PhD journey together through careful planning and decisive action. I expect you to take ownership of your progression while coming to me when you need help troubleshooting. I like to be kept up to date and will expect to see evidence of method development and data generation results on a regular basis.

Communication

I maintain transparent communication. I am known to email my team and PhD students at all hours of the day and night, though I will not expect you to do the same. We will also utilize a team chat for regular interaction during the week, and I expect you to manage your work-life balance effectively.

Meetings

PhD Students can expect scheduled meetings with me at least once per week. These meetings will be a mixture of face-to-face and via video chat or telephone. I am usually contactable for an instant response 3 or 4 days per week.

Work Patterns

The timing of work in my lab is completely flexible, and (other than attending pre-arranged meetings), I expect students to manage their own time.

Notice Period for Feedback

I need at least 1 week's notice to provide feedback on written work of up to 5,000 words.

MIBTP Project Details

Primary supervisor for:

SleeplessAI: Discovering Latent Sleep Signatures for Equitable Cardiometabolic Risk Prediction using AI on EHR and Molecular Data

See the PhD Opportunities section to see if this project is currently open for applications via MIBTP.

Please Note: The main page lists projects via BBSRC Research Theme(s) quoted and then relevant Topic(s).

SleeplessAI: Discovering Latent Sleep Signatures for Equitable Cardiometabolic Risk Prediction using AI on EHR and Molecular Data

Secondary Supervisor(s): Dr Samiul Mostafa

University of Registration: University of Birmingham

BBSRC Research Themes:

Integrated Understanding of Health (Ageing, Diet and Health)

Project Outline

Cardiometabolic diseases (CVMD)-specifically Type 2 Diabetes and cardiovascular conditions-disproportionately impact South Asian populations. This project investigates under-utilized, non-apnoea sleep abnormalities as biological mediators of CVMD. Building on extensive AI experience and preliminary data confirming sleep interventions improve metabolic markers, we aim to discover ""latent sleep signatures"" from primary care electronic health records (EHRs) to augment established risk calculators like QRISK3 and QDiabetes.

The methodology spans four phases. Phase 1 focuses on community co-design, collaborating with a Community Advisory Board to identify culturally specific, stigmatized sleep symptoms, map pharmacological confounders, and extract longitudinal EHR data for unsupervised clustering. Phase 2 employs dual-track predictive modelling to compare traditional machine learning (e.g., XGBoost) trained on structured tabular data against open-weight Large Language Models (LLMs) trained on semantic EHR narratives. We will benchmark standard versus Chain-of-Thought (CoT) fine-tuning and contrast Explainable AI (SHAP) with generative clinical reasoning to quantify incremental prognostic gain. Phase 3 ensures cross-population equity by rigorously enforcing fairness between South Asian and White European cohorts using Group Calibration, Equalized Odds, and Predictive Parity, applying both in-processing and post-processing bias mitigation. Phase 4 finalizes clinical translation via external validation using UK Biobank linkages, expert clinical auditing of AI reasoning pathways, and continuous community co-production to translate algorithmic outputs into culturally sensitive clinical tools.

The initial analysis of this project will focus on CPRD data (Data Extracted under Protocol: ID: 22_001968)

Co-supervisor on a project with