I am running WGCNA on expression data from DESeq2 (RNA-seq samples).

My pickSoftThreshold results are:

T1: power 2 gives R² = 0.963 and mean k = 8.5. Power 3 gives R² = 0.946 but mean k drops to 0.84. By power 5, mean k = 0.03 and no modules form.

T2: power 2 gives R² = 0.970 and mean k = 7.9. Power 3 gives R² = 0.970 but mean k = 0.72. Power 4 gives R² = 0.444 and mean k = 0.09.

The R² criterion is satisfied at power 2 but mean connectivity drops between power 2 and 3, making power 2 the only valid choice despite being unusually low.

Is power 2 valid here or should we use a different approach?

Software: R 4.5.2, WGCNA 1.72, unsigned TOM, minModuleSize = 30.