omicverse-microbiome-da-comparison

You can compare microbiome differential abundance results by running Wilcoxon, pyDESeq2, and ANCOM-BC on a shared two-group AnnData cohort to identify consensus hits and method-specific differences at a common FDR threshold.

To find robust differential abundance hits in sparse 16S microbiome data, apply multiple statistical methods and filter for consensus features, using FDR-aware corrections like fdr_bh or q_value to manage sparsity and compositional bias.

Yes, you need raw-count AnnData to run microbiome differential abundance method comparisons, requiring genus collapse via ov.micro.collapse_taxa and min_prevalence filtering before applying Wilcoxon, pyDESeq2, or ANCOM-BC.

The best way to choose a differential abundance method for compositional microbiome cohorts is to run Wilcoxon, pyDESeq2, and ANCOM-BC on the same contrast and evaluate which hits are robust versus method-specific under compositional bias.

You collapse microbiome features to genus level using the ov.micro.collapse_taxa function on your raw-count AnnData, then apply min_prevalence filtering before executing differential abundance tests.

Differential abundance results differ between Wilcoxon and pyDESeq2 because they handle sparsity and compositional bias differently; comparing their outputs on the same AnnData cohort helps determine which biological conclusions are method-specific.