Parkinson’s disease (PD) is a complex neurodegenerative disorder. Several risk variants and genes were identified by genetic studies and predictive disease risk models were built based on associations with common variants. We showed that singleton loss-of-function variants contribute to the genetic architecture of PD and that disease risk prediction models combining singleton and common variants improved risk models based solely on common variants. In the area of personalized medicine, more refined strategies in the construction of genetic risk scores are needed that will combine rare variant information and multiple clinical parameters with new machine learning algorithms to increase predictive power. (By Dr. Patrick May, https://jmg.bmj.com/content/early/2020/02/13/jmedgenet-2019-106316 )