sc.tl.rank_genes_groups , p value question

[ruiqiizhou]

When I use function sc.tl.rank_genes_groups(adata, 'Cell Type', method='wilcoxon', pts = True,tie_correct = True) , 'Cell Type' only contains 2 groups, I want to check adjusted p values, then I use pd.DataFrame(adata.uns['rank_genes_groups']['pvals_adj']).
It gave me 2 different columns adjusted p values. I am not sure how can I interpret these two different p values. I just wanna use p value to check interested genes differential expression in two groups. In which group is this gene expressed more significantly?
Which value I should trust？

[LuckyMD]

Hi @ruiqiizhou,

Are the p-values in both different? If you have 2 cell types and you use a general sc.tl.rank_genes_groups call, then it test cell type A vs cell type B, and cell type B vs cell type A. It's set up this way as it always tests A vs rest, B vs rest, etc. That should result in p-values for both sides of the test. They should thus be the same. If not, maybe this question is best asked on the scanpy github at scverse.