Regarding the output of the DESeq2 package, I am a bit confused. When I use the `deseq()`

function for estimation and the `lfcShrink()`

function for shrinkage estimation, I understand that the T-ratio is obtained from the ratio of the estimated coefficient to its standard error, and this T-ratio is then used to calculate the P-value and the adjusted P-value after multiple testing correction. However, why doesn't the lfcShrink() function recalculate the T-ratio, P-value, and the adjusted P-value, even though the estimated coefficients and standard errors have been altered by shrinkage?

Yes, you are correct, I meant to describe the DESeq() function. But I want to know why lfcShrink does not imply statistical inference? As I know, this should be a robust estimation method used to shrink the log2 fold change of low express genes, making it closer to the prior expected value and correcting its standard error to make the estimation more robust. So, why doesn't this imply statistical inference?

You should read the apeglm paper. I wasn't completely correct - the shrinkage methods are also meant to provide better estimates of the logFC, so if you want to rank your significant genes by the logFC, you can shrink them first to preclude low expressing genes with inflated logFC values from polluting your top genes. But it's not used for inference, which is why the t-statistics and p-values aren't affected.