Computational Analysis of the Urinary Steroid Metabolome as a Novel Non-invasive Tool to Stage Non-Alcoholic Fatty Liver Disease (NAFLD)
Ahmad Moolla, Oxford Centre for Diabetes, Endocrinology & Metabolism, University of Oxford, Churchill Hospital
- Detailed investigation of a novel non-invasive tool – measurement of the urinary steroid metabolome in point of care urine collection combined with computational machine learning – which may accurately stage NAFLD and reduce the need for liver biopsy, the current gold standard investigation.
- In-depth mathematical interrogation of the urinary steroid metabolome using generalized matrix learning vector quantisation (GMLVQ) to assist translation of this experimental technique from gas-chromatography / mass-spectrometry (GC/MS) to a more efficient, rapid and cost effective liquid-chromatography / mass-spectrometry (LC/MS) platform.
- To learn from and develop strong and innovative collaborative links with world-class, international experts in machine learning techniques from the University of Groningen, The Netherlands, and to apply these to an important medical research area.
- To allow me as a clinical trainee and early career academic to develop skills and experience in a centre of research excellence in computational analysis internationally.
- To allow me as a clinical researcher to contribute to an important research field (NAFLD) which affects up to 30% of the general population but for which diagnosis is currently complex, has morbidity and is costly, and for which there are limited treatment options.
I would like to thank the Bioscientifica Trust for this award which allowed me to visit my collaborators at the Johann Bernoulli Institute for Mathematics and Computer Science, University of Gronignen, The Netherlands to enhance my research work in non-alcoholic fatty liver disease (NAFLD).
The overall aim of this project was to investigate the utility of the urinary steroid metabolome combined with computational machine learning as a novel non-invasive tool to accurately stage NAFLD and reduce the need for liver biopsy, the current gold standard investigational tool.
To achieve this, we have interrogated urinary steroid data of >250 subjects across the spectrum of NAFLD using generalized matrix learning vector quantisation (GMLVQ) in collaboration with international experts in machine learning from the University of Groningen.
My visit to Groningen allowed me to spend good quality time with my collaborators to learn about and better understand the mathematical techniques involved. The visit was mutually beneficial for both parties as it allowed better appreciation of the challenges we are seeking to solve and detailed consideration of how to target our research work. As such, it significantly enhanced work to interrogate my dataset and to help identify a ‘NAFLD specific’ subset of the most discriminatory urinary steroids that contribute to the separation of disease states across the NAFLD spectrum. This work is on-going.
My visit also enhanced our research collaboration more generally and facilitated additional positive outcomes. For example, over the past year, I have co-supervised 2 Masters students from Groningen who have both worked on the mathematical aspects of this research. One of the students also visited Oxford and delivered a seminar on our joint work to our research group and we have similar plans to invite the second student to visit.
Grant awarded: £2,000