Researchers in Manchester are hoping to develop an improved test for growth hormone deficiency in children using machine learning and DNA analysis.
The current test requires the child to fast for up to 12 hours, then have an intravenous infusion and as many as ten blood tests over half a day to measure growth hormone production.
Dr Adam Stevens from the University of Manchester, UK, and Dr Philip Murray from Manchester University NHS Foundation Trust used machine learning to search for an improved test.
This involved using the mathematical technique of Random Forest Analysis to analyse 30,000 genes, from samples taken from 72 patients with growth hormone deficiency and 26 children without the condition.
This led to examination of approximately three million separate data points, comparing different gene patterns between the children in each group. This process highlighted 347 genes that can accurately diagnose growth hormone deficiency, then treatment with synthetic growth hormones can be prescribed.
In the Journal of Clinical Investigation – Insight recently, the team report that a test based on the discovery could be available within five years.
Dr Stevens said: "We think this is an important development in the way doctors will be able to diagnose growth hormone deficiency – a condition which causes distress to many thousands of children in the UK.
“This sort of diagnostic would not be available even a few years ago but thanks to the enormous computing power we have, and advances in genetics, it is now possible for this aspect of care to be made so much easier for patients, and the NHS."
Dr Murray added: "This study provides strong proof of concept, but before it is in a position to be adopted by the NHS, we must carry out a further validation exercise which will involve comparing our new diagnostic with the existing test."
Murray, P. G. et al. Transcriptomics and Machine Learning Predict Diagnosis and Severity of Growth Hormone Deficiency. Journal of Clinical Investigation – Insight 5 April 2018 doi: 10.1172/jci.insight.93247
Leave a Reply