So I developed an algorithm to quantify some transcripts. I want to do differential expression analysis on these transcripts (control vs treatment). Initially, I calculated log2FC and did Wilcoxon ranked test on it, and found a significantly differentially expressed transcript. Then, I did Wilcoxon directly on control vs treatment, but for the same transcript, I found no difference. So I am wondering, when we do DE analysis, what's the null hypothesis? Is it log2FC = 0 or ctrl = tret? The two null hypotheses look literally the same, but why we get different results eventually?