Overview
This highly interdisciplinary studentship brings together the expertise of the Genomics and Computational Biology Group with the Embedded and Intelligent Systems Laboratory, also leveraging data from the UK Household Longitudinal Study held from Institute for Social and Economic Research (ISER), which could in turn enable a significant scientific impact potential for this timely research endeavour.
Therefore, we are looking for a highly motivated and interdisciplinary minded student, who has an excellent computer science, bioinformatics, electronics or related UG/MSc degree and is keen to work on relevant research in the area of design deep learning architectures and algorithms for age prediction based on DNA methylation data.
The project
Aging and health is a supremely timely topic as life expectancies worldwide are higher than ever and non-communicable causes of death now predominate for a majority of the world's population.
DNA methylation profiles are of great interest because they are an epigenetic measure related to gene expression and its regulation but more stable and easier to analyse than gene expression itself or other epigenetic marks.
One topic of particularly wide interest currently is epigenetic change with age, both as a way of accurately aging unknown DNA samples and as a way of understanding some aspects of biological aging and health and environmental effects on it. Deep learning-based machine learning technologies have demonstrated their increasing usefulness as a powerful tool in pattern recognition and demonstrated outstanding performance (predominantly outperforming traditional non-artificial intelligence bases approaches) in visual object identification, speech recognition, biomedical signal analysis and other fields requiring hierarchical analysis of input data.
Inspired by these examples, the aim of this project is to explore if deep convolutional neural networks (CNNs) architectures can be employed to generate similarly spectacular results for age-specific DNA methylation patterns to generate an accurate model for the prediction of chronological age and tissue classification using data from blood samples.
The main anticipated outcome for this project is that the proposed deep CNN architecture would simultaneously and automatically producing the set of features engineered from the raw DNA methylation data and bring age prediction and tissue classification to a new level, which could be potentially used with a combination of wearable sensors and deep learning technologies for Health Risks Assessment (HRA) applications involving continuous health risk evaluation and real-time feedback to patients and care providers in future.