Abstract Background Atrial fibrillation (AF) is the most common cardiac arrhythmia in athletes. Despite a lower burden of classical cardiovascular risk factors, the prevalence of AF in athletes is comparable to that in the general population of similar age (~13%, mean age 65.5), suggesting that exercise-induced AF may have different underlying mechanisms. The prevalence of AF in thoroughbred (TB) and standardbred (STB) racehorses is also relatively high (~5%), suggesting exercise as an independent risk factor. Due to selective breeding and complete control of pedigrees, racehorses are well-suited for genetic studies. Common and rare genetic variation explain a substantial part of AF variance, but no study has identified genetic variants associated with exercise-induced AF in any species. Objective In this project, we aimed to use the racehorse as a model to identify genetic variants associated with exercise-induced AF, and in a future study, possibly replicate the results in human athletes. Here we present preliminary results from the genome-wide association study (GWAS) of AF in racehorses. Methods We collected blood or hair samples from 181 STB and TB racehorses in which 71 had paroxysmal or persistent AF verified by ECG. We extracted DNA and performed whole-genome genotyping using a commercial equine genotyping platform (670k markers). We phased and imputed un-genotyped markers and increased the sample size of the control group, using publicly available equine data from similar breeds from the European Nucleotide Archive. We conducted a GWAS of 7.85 million markers in 71 cases and 296 controls using PLINK2 with covariates adjusted for population stratification, batch and sex. Results We identified one significant locus downstream of the protocadherin gene PCDH18 associated with AF in racehorses. The same locus has been associated with the ECG-trait P-wave terminal force in humans. Changes in P-wave terminal force is associated with increased atrial size, delayed interatrial conduction, left ventricular fibrosis and left atrial function, and considered a good predictor of AF occurrence. Several non-significant but interesting signals were also identified including the genes SLC6A15 (amino acid transport), MYBPH (myosin binding), MYF6 and MYOG (myoblast differentiation). Conclusion In these preliminary results, we have identified a new genetic locus associated with AF in racehorses. In addition, we have identified several candidate genes for further studies to elucidate the genetic background of exercise-induced AF. We aim to replicate our findings in a population of human athletes. Our results may aid the development of risk prediction tools for exercise-induced AF and new drugs for tailored treatment of AF in general.