Understanding dynamic steroid biosynthesis in health and disease through machine learning in the space of mechanistic models

 

Lead Academic Co-Investigators Centre Fellow(s) Project title 
Peter Tino

Georgina Russell (BIRM)

Thomas Upton (BRISTOL)

Stafford Lightman (BRISTOL)

Eder Zavala (CSMQB)

Yuan Shen (BIRM)

Krasimira Tsaneva-Atanasova (CEMPS)

Diane Fraser

Xinyue Chen

Understanding dynamic steroid biosynthesis in health and disease through machine learning in the space of mechanistic models

 

Lay summary:

Hormones, including the stress hormone cortisol, are released in rhythmic patterns. This means that there are variations in normal levels across time and between individuals. Consequently, traditional, one-off hormone measurements are extremely difficult to interpret, which can lead to delay in diagnosis and treatment.  

To address this problem, we need to understand hormone behaviour over the day. We will explicitly model biological mechanisms involved in hormone dynamics. Given hormone measurements taken from patients and healthy volunteers over 24 hours, we will investigate how the data could be explained and classified with the help of our modelling. This will be achieved via the development of a mathematical framework that will enable us to quantify to what degree the data from a particular subject “is explainable” by models related to normal or pathological cases. In other words, to what degree (quantified as probability) the subject should be diagnosed as a normal or pathological case.

The primary effect will be to benefit patients. Diagnosis will be understandable in terms of the underlying biology, faster, more reliable and convenient. Treatment will be easier to monitor and tailored to the individual. The overall effect will be to reduce the burden on our healthcare system.

Google+