Researchers from the University of Surrey, have developed a machine learning algorithm that can accurately estimate an individual’s circadian phase using targeted metabolomics data from one or two optimally timed blood samples.
The related paper Machine learning estimation of human body time using metabolomic profiling[1] was recently published on 24th April 2023.
It discussed an algorithm that outperforms previously published approaches and provides a cheap and practical method for potential circadian applications in the clinic.
Why Care About Estimating an Individual’s Body Clock
Circadian rhythms influence physiology, metabolism, and molecular processes in the human body.
It has significant implications for optimizing behavior, diagnostic sampling, medical treatment, and the treatment of circadian rhythm disorders.
Individual circadian phase can vary greatly between individuals and is influenced by a number of factors such as age, sex, genetic makeup, and environmental factors.
Thus, accurate estimation of an individual’s circadian phase is vital for personalized optimization of behavior and medical treatment.
For instance, an accurate estimation of an individual’s circadian phase could help optimize the timing of drug administration or the scheduling of medical procedures for maximum efficacy and minimal side effects.
It could also aid in the optimization of individual behavior, such as sleep, meals, and physical activity.
Partial Least Squares Regression (PLSR) Machine Learning Approach
The researchers developed a PLSR machine learning approach that uses plasma-derived metabolomics data in one or more samples to estimate dim light melatonin onset (DLMO) as a proxy for the circadian phase of the human body.
DLMO is a commonly used marker for the circadian phase as it indicates the time of day at which the melatonin concentration in the blood starts to increase.
Researchers optimized their approach for either women or men under entrained conditions, which means their subjects followed a consistent sleep-wake cycle over several days to mimic real-life conditions.
Results demonstrated that the PLSR machine learning approach performed equally well or better than existing approaches that utilize more labor-intensive RNA sequencing-based methods.
The PLSR machine learning approach relies on identifying a set of metabolites in the blood that are associated with circadian rhythms.
These metabolites are then used as inputs for the PLSR algorithm to predict the circadian phase based on the DLMO marker, the approach’s design ensures that it remains practical for use in clinical settings.
The researchers’ approach provides a cheap and practical method to predict an individual’s circadian phase accurately.
It can help clinicians identify the optimal time for diagnostic sampling and medical treatment based on an individual’s circadian phase, leading to improved treatment outcomes and reduced side effects.
The approach’s reliance on targeted metabolomics data from one or more blood samples eliminates the need for more invasive and time-consuming RNA sequencing-based methods.
Future Implications of the Research in Circadian Rythm Diagnosis
For individual timing of medical treatment or diagnosis of circadian disorders, it is important to develop practical methods that allow the estimation of individual circadian phases in the clinic.
The machine learning algorithm developed by the researchers provides a cheap and practical method for potential circadian applications in the clinic after appropriate validation.
Researchers hope that their approach could be used to develop a circadian phase estimation tool that could be incorporated into electronic medical records, allowing clinicians to quickly and easily estimate an individual’s circadian phase and make personalized treatment decisions.
The estimation of circadian body time using blood-targeted metabolomics requires further validation in shift work and other real-world conditions.
For now, it offers a robust and feasible technique with relatively high accuracy to aid personalized optimization of behavior and clinical treatment after appropriate validation in patient populations.
References
- Tom Woelders, Victoria L. Revell et al., ‘Machine learning estimation of human body time using metabolomic profiling‘, Proceedings of the National Academy of Sciences (PNAS), 24 April 2023, https://www.pnas.org/doi/10.1073/pnas.2212685120[↩]
