Genome-wide Association Study of Liking for Several Types of Physical Activity in the UK Biobank and Two Replication Cohorts

Introduction A lack of physical activity (PA) is one of the most pressing health issues today. Our individual propensity for PA is influenced by genetic factors. Stated liking of different PA types may help capture additional and informative dimensions of PA behavior genetics. Methods In over 157,000 individuals from the UK Biobank, we performed genome-wide association studies of five items assessing the liking of different PA types, plus an additional derived trait of overall PA-liking. We attempted to replicate significant associations in the Netherlands Twin Register (NTR) and TwinsUK. Additionally, polygenic scores (PGS) were trained in the UK Biobank for each PA-liking item and for self-reported PA behavior, and tested for association with PA in the NTR. Results We identified a total of 19 unique significant loci across all five PA-liking items and the overall PA-liking trait, and these showed strong directional consistency in the replication cohorts. Four of these loci were previously identified for PA behavior, including CADM2, which was associated with three PA-liking items. The PA-liking items were genetically correlated with self-reported (r(g) = 0.38-0.80) and accelerometer (r(g) = 0.26-0.49) PA measures, and with a wide range of health-related traits. Each PA-liking PGS significantly predicted the same PA-liking item in NTR. The PGS of liking for going to the gym predicted PA behavior in the NTR (r(2) = 0.40%) nearly as well as a PGS based on self-reported PA behavior (r(2) = 0.42%). Combining the two PGS into a single model increased the r(2) to 0.59%, suggesting that PA-liking captures distinct and relevant dimensions of PA behavior. Conclusions We have identified the first loci associated with PA-liking and extended our understanding of the genetic basis of PA behavior.