Why do FindMarkers DE p-values vary significantly between V4 and V5?

[mauritsunkel]

Hi,

In Seurat V5 news, it was stated that the DE pseudocount is changed, now performed at the group level rather than cell level (#8462). (Next to min.pct defaults having gone from 0.1 to 0.01 and logfc.threshold from 0.25 to 0.1)
There is no note about p-value calculation having changed at all.

In V5, especially very lowly expressed genes in either groups (pct.1 || pct.2 < 0.01) now have substantially higher log2FC, however, they also have substantially lower p-val(-adj). In V4, I would filter out a relevant portion of these lowly expressed genes based on having high p-value.

I would like to run the new V5 estimations for log2FC, yet keep the V4 calculation for p-values, as expected.
Could it be explained/noted why p-values are changed in V5 from V4 as well?

Best,
Maus