Getting less DE genes when incorporating all samples

Question

[shome]

When I am using 18 samples, I am getting upregulated genes as 2 and downregulated genes as 0.

Now, when I remove four samples, I am getting upregulated genes as 2000 and downregulated genes as 1000, using DESEq2.

What should be the most appropriate way to deal with this?

[swbarnes2]

Well, what have you looked at? Are those 4 samples outliers in the PCA? Is there reason to think there is an experimental reason for them to be bad? Or are you artificially removing accurate, but inconvenient samples to get the desired results?

[jv]

I totally agree with the above comments.

Also, see the following in the DESeq2 tutorial: If I have multiple groups, should I run all together or split into pairs of groups?

[ATpoint]

Generally, for diagnosis you need to show code, experimental design and diagnostic plots such as PCA and MA-plots. Without that it's hard.