Question about FindMarkers avg_logFC and how to get all features from an assay
0
0
Entering edit mode
3.6 years ago
compuTE ▴ 140

I have added a new assay to my seurat object and I want to test for differential expression of all these features by a condition in a metadata column (which for this example can take the values of "condition1" or "condition2"). My command looks like:

dea <- FindMarkers(obj, ident.1 = "condition1", ident.2 = "condition2", 
                   group.by = "condition", assay = "another_assay", features = rownames(obj@assays$another_assay),
                   test.use = "wilcox",  min.pct = 0, logfc.threshold = 0, base = 2)

This does not return the test results for all the features in the assay :-/ I know the default for features = is all genes in the assay but i was not getting all the features, so i specifically put the rownames of the assay to specify it, but i still get less than I would expect.

I have also tried logfc.threshold = -Inf, I get the same output.

I have tried setting min.diff.pct = 0, I get even fewer features.


In addition to this, I have a question regarding the output I do get. The documentation says:

avg_logFC: log fold-chage of the average expression between the two groups.

For a given feature, I'm getting an avg_log2FC of 8.5 . Just to verify, I did:

feature1 <- data.frame(V1=GetAssayData(obj, assay = "another_assay")["feature1",])
feature1 <- merge(feature1, data.frame(condition=obj$condition), by='row.names')

avg_cond1 <- mean(feature1[which(feature1$condition == "condition1"), "V1"])
avg_cond2 <- mean(feature1[which(feature1$condition == "condition2"), "V1"])

log2(avg_cond1 / avg_cond2)

This does not equal 8.5

Any idea on what I'm doing wrong?

I'm running R 4.0.2 with Seurat 4.0.0


UPDATE: This gitHub issue solves my second question. However, I have a follow up question:

In their avg_logFC calculation they calculate the mean of e^x - 1 (where x is the expression values for a gene) and then log it (line 176/7).

My question is: Why do they calculate the mean of e^x - 1 instead of just the mean of x?

seurat singlecell Seurat • 2.3k views
ADD COMMENT

Login before adding your answer.

Traffic: 2207 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6