Question

FindAllMarkers pct1 and pct2

0

Entering edit mode

5.0 years ago

Morris_Chair ▴ 370

Hello guys,

I have a doubt about the output of FindAllMarkers, I know that pct.1 is the percentage of features "A" in cluster 1 but I am not sure about pct.2,

Is it the percentage of that features "A" in all the other clusters ? since I am comparing one vs the rest?

object.markers <- FindAllMarkers(object = myobject,only.pos = TRUE, min.pct = 0.25, thresh.use = 0.25)

Thank you

RNA-Seq Seurat • 14k views

ADD COMMENT • link updated 2.7 years ago by jared.andrews07 ★ 18k • written 5.0 years ago by Morris_Chair ▴ 370

score 6 · Accepted Answer · 2019-11-28

6

Entering edit mode

5.0 years ago

jared.andrews07 ★ 18k

I'm not really following what you're talking about with pct1 and pct2, but the min.pct parameter just determines the fraction of cells in either group the feature has to be expressed in to be included in the analysis. It's meant to remove lowly expressed genes. So in your case, that'd mean for a gene to be included in the differential expression for cluster 1 (or 2, 3, etc) versus the other clusters, at least 25% of cells in cluster 1 or all other clusters must express the gene.

ADD COMMENT • link 5.0 years ago by jared.andrews07 ★ 18k

0

Entering edit mode

Hi Jared. Can you explain what does only.pos = TRUE argument does. I have been reading around this but couldn't find a concrete explanation for this argument. It was used in the following seurat issues: 1, 2, 3. A similar question here went unanswered for years now.

ADD REPLY • link 2.7 years ago by rohitsatyam102 ▴ 920

0

Entering edit mode

See ?FindMarkers.

only.pos
Only return positive markers (FALSE by default)

ADD REPLY • link 2.7 years ago by jared.andrews07 ★ 18k

0

Entering edit mode

So it reports Marker genes with positive avg_log2FC right?

ADD REPLY • link 2.7 years ago by rohitsatyam102 ▴ 920

1

Entering edit mode

That is correct.

ADD REPLY • link 2.7 years ago by jared.andrews07 ★ 18k

score 6 · Accepted Answer · 2020-07-30

FindAllMarkers function helps to identify gene markers for each cluster relative to all other clusters. The expression for a given gene among cells in a given cluster is compared against the expression of that gene among cells in all other clusters. In the output, pct.1 is the percentage of cells in the cluster where the gene is detected, while pct.2 is the percentage of cells on average in all the other clusters where the gene is detected. A gene to be considered as an IDEAL cluster marker is expected to be expressed exclusively in that cluster and silenced in all others and thus pct.1 will be more towards 1 and pct.2 towards 0.