![]() ![]() This allowed factors such as diet, exercise and intake of pharmaceuticals to be evaluated for their influence on longevity. DNA methylation clocks have then been used to assess behavioral lifestyle factors for their effect on biological age ( Quach et al., 2017). Perhaps the best described “aging clock” is based on epigenetic age via DNA methylation ( Bocklandt et al., 2012 Hannum et al., 2013 Horvath, 2013). A composite biomarker predictor has also been developed, utilizing 18 biomarkers over multiple organ systems in young adults ( Belsky et al., 2015). Right panel is the normalization of the prediction based on a priori knowledge of the participant’s ages, resulting in r = 0.94, p < 0.01, and an RMSE of 7.54 yearsĪ variety of biological age predictors have been generated already, using parameters such as telomere length, gene expression profiles, or metabolomics (reviewed by Jylhävä et al., 2017). Left panel is direct prediction on the raw accelerometer data form the validation dataset, with r = 0.75, p < 0.01, and an RMSE of 13.58 years. (G) The prediction of the models for the validation dataset. Strength of each predictor is interpreted from the models percent increase of mean standard error (Percent Inc. (F) The strength of each predictor (variance or maximum intensity), by day and hour in the dataset, for the random forest machine learning model. The maximum intensity and variance of readings per hour over the seven days were used as input for machine learning to predict age of the individual. (E) Example of hourly maximum intensity (top panels) and hourly variance (bottom panels) for the 18 and 80 year olds depicted in panel D. ![]() The data covers seven days of readings (top panels), and a single day contains readings from the moment the individual attached the device in the morning until they removed it in the evening (bottom panels). (D) Example of accelerometer readings in the data for a typical 18 year old (left hand panels) and typical 80 year old (right hand panels). Left panel is the 2003–2004 dataset and right panel is the 2005–2006 dataset. (C) The distribution of ages and total counts of individuals for each dataset, following data quality filtering steps. ![]() The 2005–2006 dataset was used for validation and exploration of associations with accelerated or decelerated biological aging. The 2003–2004 dataset was used for model building and validation. The two publically available NHANES datasets that included accelerometer data were used. (B) Schematic showing machine learning strategy to build age predictors from wearable devices that measure accelerometer readings. Genetics, drugs, and nutrition are able to promote healthy aging, or, in the case of drugs and nutrition, can promote healthy aging and thereby improved locomotive capacity (blue figures). (A) Schematic representing the typical aging process (black figures), accompanied by reduced locomotive capacity. Machine learning to predict age from wearable device movement data. Our work demonstrates how a biological aging score based on relative mobility can be accessible to the wider public and can potentially be used to identify and determine efficacy of geroprotective interventions.įIGURE 1. We show that doxazosin extends healthspan and lifespan in C. We additionally identified one FDA-approved drug associated with decelerated biological aging: the alpha-blocker doxazosin. A number of nutritional components peak in their association to decelerated aging later in life, including fiber, magnesium, and vitamin E. We further searched for nutritional or pharmacological compounds that associate with decelerated aging according to our model. We found that accelerated biological aging from our “MoveAge” predictor is associated with higher all-cause mortality. Here we used NHANES physical activity accelerometer data from a wearable device and machine-learning algorithms to derive biological age predictions for individuals based on their movement patterns. ![]() However, the ability to measure the biological aging process in individuals, which is necessary to test for efficacy of these interventions, remains largely inaccessible to the general public. Intervening in aging processes is hypothesized to extend healthy years of life and treat age-related disease, thereby providing great benefit to society. 3NemaLife Inc., Lubbock, TX, United States.2Department of Chemical Engineering, Texas Tech University, Lubbock, TX, United States.1Laboratory Genetic Metabolic Diseases, Amsterdam Gastroenterology, Endocrinology, and Metabolism, Amsterdam Cardiovascular Sciences, Amsterdam UMC, University of Amsterdam, Amsterdam, Netherlands. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |