Thanks for developing the MaAsLin2 that enables incorporating random effects in the models. To reduce the number of multiple testing, MaAsLin2 allows filtering uninformative features that have low counts and/or low prevalence. However, the filtering thresholds are somewhat arbitrary, often determined based on users’ knowledge or preference. The differential gene expression analysis software, DESeq2, has a function for automatic independent filtering of features that maximizes the number of genes found at a user-specified target FDR. Is it possible to implement a similar function in MaAsLin2?