Incorporating prior information and borrowing information in high-dimensional sparse regression using the horseshoe and variational Bayes.