Closed danielcgingerich closed 3 years ago
How did you process the data upstream of doing the differential testing? Can you provide the code, and output of sessionInfo()
?
Any update here?
Hey @timoast , sorry for not getting back to you. I have been investigating this today, and it seems that signac was using the expm1
function on the 'data' slot in its fold change calculation, thus treating the data as normalized scRNA counts. Not sure why this was happening, as I ran the same code again today and got the correct results. I will post my code when I get home later today
This was due to a default in Seurat (designed for log-normalized data), that we now override in Signac 1.3.0 (see release notes)
@timoast Interesting. My signac version is 1.1.1.9010 still, but it worked correctly this time. I dont think I changed anything or updated anything since the last time I ran it Edit: I havent updated it since march 25
Not sure, maybe you set slot="counts"
in FindMarkers()
? Anyway, should be fixed in the current release
12 Disease + 12 control samples were integrated and clustered. In each cluster, differential accessibility was performed for disease cells vs control cells. The results are very odd for a two reasons.
1) most differentially accessible peaks (DAPs) have positive log fold change. I expect a relatively even distribution.
2) For each group of DAPs - more accessible vs less accessible - I looked at the average percent expression of the peaks (mean(pct.1, pct.2)) across all cells that were tested. Less accessible DAPs have very low percent expression compared to more accessible DAPs.
This kind of suggests to me that the results are still being confounded. It seems that diseased samples were sequenced at much higher depth than normal samples.
FindMarkers()
was done using 'LR' test andlatent.vars = 'age' + 'sex' + 'pmi' + 'peak_region_fragments'
. However, these variables only explain about 2% of variation of the first LSI component.This does not appear to be an issue with
FindMarkers
. I did the LR test manually and obtained the same results.My question is, are these results normal? Should I be concerned?