Machine learning approaches to predict age from accelerometer records of physical activity at biobank scale.
Machine learning approaches to predict age from accelerometer records of physical activity at biobank scale.
复制标题
DOI：
10.1371/journal.pdig.0000176
复制
发表时间：
2023-01
期刊：
影响因子：
--
通讯作者：
中科院分区：
文献类型：
--
作者：
体育活动可以提高生活质量，预防与年龄有关的疾病。随着年龄的增长，体力活动趋于减少，增加了老年人对疾病的脆弱性。在下文中，我们训练了一个神经网络来预测来自英国生物银行的115，456个为期一周的100 Hz手腕加速度计记录的年龄（平均绝对误差= 3.7±0.2岁），使用各种数据结构来捕捉现实世界活动的复杂性。我们通过将原始频率数据预处理为2，271个标量特征，113个时间序列和4个图像来实现这一性能。我们将参与者的加速老化定义为预测年龄大于实际年龄，并确定了与新表型相关的遗传和环境暴露因素。我们对加速老化表型进行了全基因组关联以估计其遗传率（h_g2 = 12.3±0.9%），并鉴定了与第六染色体上的组蛋白和嗅觉簇中的基因非常接近的10个单核苷酸多态性（例如HIST1H1C，OR5V1）。同样，我们确定了与加速衰老相关的生物标志物（如血压）、临床表型（如胸痛）、疾病（如高血压）、环境（如吸烟）和社会经济（如收入和教育）变量。体力活动导致的生物学年龄是一个复杂的表型，与遗传和非遗传因素有关。体育活动可以提高生活质量，也是流行的年龄相关疾病和结果的重要保护因素，如糖尿病和死亡率。随着年龄的增长，体力活动趋于减少，增加了老年人对疾病的脆弱性。从数字健康设备测量的身体活动是否可以预测一个人的生物年龄？与实足年龄（出生后所经过的时间）相比，生物年龄是随着时间的推移而累积的生物变化的指标，这些变化被假设为与年龄有关的疾病的一个因果因素。在下文中，我们训练了机器学习模型，以预测来自英国生物银行参与者的115，456个为期一周的手腕加速度计记录的年龄。然后，我们发现了与加速衰老相关的遗传、环境和行为因素，即生物学年龄和实足年龄之间的差异，为我们新预测因子的生物学可解释性提供了证据。如果是可逆的，将复杂的身体活动总结为生物年龄预测器可能是实时观察预防措施效果的一种方法。
Physical activity improves quality of life and protects against age-related diseases. With age, physical activity tends to decrease, increasing vulnerability to disease in the elderly. In the following, we trained a neural network to predict age from 115,456 one week-long 100Hz wrist accelerometer recordings from the UK Biobank (mean absolute error = 3.7±0.2 years), using a variety of data structures to capture the complexity of real-world activity. We achieved this performance by preprocessing the raw frequency data as 2,271 scalar features, 113 time series, and four images. We defined accelerated aging for a participant as being predicted older than one’s actual age and identified both genetic and environmental exposure factors associated with the new phenotype. We performed a genome wide association on the accelerated aging phenotypes to estimate its heritability (h_g2 = 12.3±0.9%) and identified ten single nucleotide polymorphisms in close proximity to genes in a histone and olfactory cluster on chromosome six (e.g HIST1H1C, OR5V1). Similarly, we identified biomarkers (e.g blood pressure), clinical phenotypes (e.g chest pain), diseases (e.g hypertension), environmental (e.g smoking), and socioeconomic (e.g income and education) variables associated with accelerated aging. Physical activity-derived biological age is a complex phenotype associated with both genetic and non-genetic factors. Physical activity improves quality of life and is also an important protective factor for prevalent age-related diseases and outcomes, such as diabetes and mortality. With age, physical activity tends to decrease, increasing vulnerability to disease in the elderly. Does physical activity measured from digital health devices predict one’s biological age? Biological age, as contrast to chronological age (the time that has elapsed since birth), is an indicator of the biological changes that accrue through time that are hypothesized to be one causal factor for age-related diseases. In the following, we trained machine learning models to predict age from 115,456 one week-long wrist accelerometer recordings from participants of the UK Biobank. We then found genetic, environmental, and behavioral factors associated with accelerated age, the difference between biological and chronological age, adding to the evidence of the biological plausibility of our new predictor. If reversable, summarizing complex physical activity into a biological age predictor may be a way of observing the effect of preventative efforts in real-time.