Quantitative approaches to energy and glucose homeostasis: machine learning and modelling for precision understanding and prediction

Abstract

<jats:p>Obesity is a major global public health problem. Understanding how energy homeostasis is regulated, and can become dysregulated, is crucial for developing new treatments for obesity. Detailed recording of individual behaviour and new imaging modalities offer the prospect of medically relevant models of energy homeostasis that are both understandable and individually predictive. The profusion of data from these sources has led to an interest in applying machine learning techniques to gain insight from these large, relatively unstructured datasets. We review both physiological models and machine learning results across a diverse range of applications in energy homeostasis, and highlight how modelling and machine learning can work together to improve predictive ability. We collect quantitative details in a comprehensive mathematical supplement. We also discuss the prospects of forecasting homeostatic behaviour and stress the importance of characterizing stochasticity within and between individuals in order to provide practical, tailored forecasts and guidance to combat the spread of obesity.</jats:p>

The homeostatic dynamics of feeding behaviour identify novel mechanisms of anorectic agents

“…Doing so would allow us to disambiguate between the effects due to stomach fullness (e.g., ghrelin release or mechanoreception) and effects due to the small or large intestine. Examples of suitable physiological models are reviewed in [63], and more continue to emerge, such as a recently developed model for ghrelin dynamics [64]. To accurately parametrise such models would require larger-scale data collection but could potentially yield new insights into the ways in which feeding behaviour is controlled.…”

Uncovering personalized glucose responses and circadian rhythms from multiple wearable biosensors with Bayesian dynamical modeling

“…There are multiple possible approaches to modeling multimodal data such as we collected, and the particular structure of the glucose model has often been dictated by the data available and by the stated goal. 34 , 35 , 36 Models based on differential equations range from simple, minimal models 37 to mechanistically detailed descriptions that include more variables, more spatial compartments, and dozens of additional parameters. 38 Inspired by these more complex models, there are many possible extensions that could be added to our glucose model, such as glucose absorption rates of mixed meals due to food content in carbohydrates, but also fat, fiber, and protein contents, which are known to slow down nutrient absorption, 14 although the addition of meal-specific response shapes would effectively double the number of meal-related parameters.…”

“…Regarding the analysis of wearable data streams, a diverse range of glucose models have been proposed over the last decades, 34 , 35 , 36 ranging from minimal models 37 to more detailed simulators with dozens of parameters 38 and neural networks. 39 , 40 Recent efforts have also attempted to utilize additional multimodal wearable signals to either improve glucose forecasting or provide more accessible proxies for glucose without using CGMs.…”

Uncovering personalised glucose responses and circadian rhythms from multiple wearable biosensors with Bayesian dynamical modelling

“…Many different models of glucose dynamics have been proposed over the last several decades [44][45][46] , ranging from minimal models 47 to more detailed simulators with dozens of parameters 48 . There are many possible extensions that could be added to the glucose model.…”