Which plan is more reasonable, gsea on total dataset , on subset of total dataset with DEG or preRanked gsea?

Question

[Shamim Sarhadi]

Hi Dears I have an expressionSet including case and control samples, I did DEG analysis with limma, now I want to do gsea analysis for 2 phenotype , but I am hesitate about choosing above mentioned approachs and which one is more reasonable

1. gsea on total dataset including case and control
2. gsea on subset of total dataset with DEG between case and control that come from limma toptable output
3. preRanked gsea (including a ranked gene list )

Thanks in advance