Regarding the multiple testing correction approach, here we have used Bonferroni correction using the total number of genes tested. In general, both single-variant and PrediXcan analyses will be performed; thus the question that arises is how to address the issue of multiple testing adjustment. The prior probability for a SNP to be causal is much smaller than the prior probability of causality for a gene so it would not be fair to subject SNP tests and gene-based tests to the same level of adjustment. Since we are presenting only gene-based results in our application and given the highly conservative nature of Bonferroni correction, there is no need to further adjust our results. A more conservative approach would be to divide the significance threshold used by a factor of two for the multiple testing using gene-based and SNP-based approaches.
