Evolution of Enriched-Phenotypes for Health Exploration and Discovery
Body composition studies based on large medical imaging databases with tens of thousands of subjects, such as the UK Biobank, require automated and robust tools for preprocessing and segmentation. We have developed deep-learning-based segmentation models capable of measuring the volume, shape, fat and iron content of multiple organs and tissues (including abdominal organs, adipose tissue, bones, and muscles). These methods enable robust measurement of image-derived phenotypes from population-scale studies, facilitating genetic and phenotypic studies to be conducted in health and disease.
This work is funded by Calico LLC.
For further information contact:
- Professor Jimmy Bell at [email protected]
- Dr Brandon Whitcher at [email protected]
- Professor Louise Thomas at [email protected]
Prediction of Body Composition by Deep-Learning methodology
While we can accurate phenotyping of body composition by MRI is now routine, access to MRI scanners is expensive and, hence rarely included in global population-based studies. To overcome this we are developing methods based on gold-standard MRI measurements to accurately predict body composition parameters including visceral and subcutaneous adipose tissue, liver fat and muscle content based on simple and low-cost measurements.
For further information contact:
- Professor Jimmy Bell at [email protected]
- Dr Brandon Whitcher at [email protected]
- Professor Louise Thomas at [email protected]
Portability of Genetic Risk Scores to Underrepresented Populations
The goal of the research is to port genetic risk prediction models for some of the most common diseases, trained in European populations, to Africans. We build on our track record of large-scale implementation of polygenic risk scores, collaborations with the Breast Cancer Association Consortium (BCAC) and our involvement in developing polygenic risk scores for lipid traits in African populations. Research conducted for this work will have an impact on our ability to translate European-trained models for the prediction of genetic risk not only for Africans but also for other underrepresented populations. Our research outputs will make the fruits of human genome research more equitable, diverse and inclusive.
For further information contact Dr Manuel Corpas at [email protected].
AI-driven sustainable solutions for food and water waste
As emphasised in the “AI for Good Global Summit (2017)”, the development of Artificial Intelligence (AI) driven solutions is vital in addressing global challenges related to hunger, health, education, etc. – henceforth, a vital aspect of this research is using AI to reduce and manage food waste effectively by converting them into value-added products. The project aims to drive on data-driven decisions to establish an innovative, circular food model using bio-resources recovered from food waste to grow food for consumption.
Part of this work (the Cavendish Living Lab) is funded by Quintin Hogg Trust.
For further information contact Dr Dipankar Sengupta at [email protected].
Nutritional epidemiology and disease outcomes
Robust analysis and interpretation of nutritional and dietary data is essential to inform public health advice and prevent confusion/ poor messaging (across public health and sports nutrition contexts). This work uses epidemiological data to clarify and refine public health messaging and ensure it is appropriate for the desired setting e.g., the use of BMI for athletes; energy requirements not translated to reflect altered needs based on clinical/sports requirements; and the need for reactive personalisation of advice.
For further information contact Dr Claire Robertson at [email protected].
Precision Medicine: Algorithms for Patient Diagnostic and Prognostic (longitudinal analysis) Applications
The primary objective of this project is to apply Data Science, encompassing Artificial Intelligence, Machine Learning, and Statistics, to enhance our comprehension of diseases mainly cardiovascular issues, cancer, and cognitive impairment. We are exploring diverse heterogeneous clinical datasets, including omics and related health data (for example, health apps), in order to identify patient sub-cohorts, factors causing these differences, and mapping corresponding genotype-phenotype relationships. Consequently, this is paving the way for the designing of novel algorithms and the creation of computational tools aimed at facilitating early diagnosis or prognosis for patients. One of the key developments we are working on is a Jacobian-based temporal algorithm that can aid in patient prognosis.
For further information contact Dr Dipankar Sengupta at [email protected].
Quantum Biology
In recent years it has become clear that there is a need to reassess some of our understanding of some fundamental biological processes and where quantum effects may play a fundamental role. We are currently developing and applying novel methodologies to determine the potential role of these quantum phenomena, especially in mitochondria, and how they may pertain to important biological and clinical phenomena.
For further information contact Prof. Jimmy Bell at [email protected] or Dr Rhys Mould at [email protected].
Improving the quality and safety of medicinal plant products
Research on the quality, safety and efficacy of medicinal plants and medicinal plant products. A variety of chemical and biological tests are used to investigate the intrinsic properties of medicinal plants. As well as laboratory studies, qualitative methods are used to get a better understanding of how and why certain plants have been used traditionally by diverse minority groups. The data is used to help improve the quality and safety of products available on the UK market and to further investigate the potential therapeutic use of these materials either as whole extracts or as single active compounds.
For further information contact Dr Anthony Booker at [email protected].