A machine learning-derived polygenic risk score reveals that healthy lifestyle counteracts obesity-related mortality.

Abstract

Obesity is influenced by genetic predisposition and lifestyle. The associations among genetic susceptibility to obesity, lifestyle, and all-cause mortality remain unexplored. Our goal is to develop and validate a machine learning model to assess the genetic risk of obesity and examine its association with lifestyle and all-cause mortality. We integrated genetic data from 482,700 UK Biobank participants and 8,607 Nanfang Hospital participants to create and validate a stacked machine learning model, which generates an obesity-related polygenic risk score (OPRS), to evaluate the relationships among genetic risk of obesity, lifestyle, and all-cause mortality. The model achieved area under the receiver operating characteristic curve values of 0.621, 0.616, and 0.565 for the training, internal, and external test cohorts, respectively. A high OPRS is associated with increased all-cause mortality, with a linear relationship observed among individuals with normal weight or overweight. Among individuals with a high genetic risk of obesity, adhering to four healthy lifestyle factors reduced the risk of all-cause mortality by 59% compared to those who did not. Thus, high genetic risk of obesity is associated with higher risk of all-cause mortality, but a healthy lifestyle mitigates this risk.