Discussion of "Gene hunting with hidden Markov model knockoffs'

Abstract

We congratulate the authors on this interesting paper that tackles the difficult problem of extending the work of Candès et al. (2018) to structured predictors possessing a particular type of dependence. Motivated by the aim of making the knockoff construction applicable to genome-wide association studies, Sesia et al. (2019) propose using a hidden Markov model to generate knockoffs that capture patterns of DNA variation, following the work of Scheet & Stephens (2006).