Markers

Purpose

Marker detection identifies genes that are enriched in one group of cells compared with another group.

In Seurat, marker detection is usually done with:

• FindMarkers()
• FindAllMarkers()

The same functions are used for different questions. The key is to define what groups are being compared.

Common Marker Questions

Question	Typical comparison
Cluster markers	one cluster vs other clusters
Cell type markers	one annotated cell type vs other cell types
Condition markers	one condition vs another condition within the same cell type or cluster

Do not mix all cells together for condition comparison if the real question is cell-type-specific differential expression.

Choose Assay

For marker detection, use the biologically meaningful expression assay.

Common choices:

Workflow	Common assay for markers
Classic workflow	"RNA"
SCT workflow	"SCT"
CCA/SCT integration	usually "SCT" or "RNA", not "integrated"
Harmony integration	usually "SCT" or "RNA"

The integrated assay is mainly for alignment and clustering. It is usually not the assay used for marker testing.

Set the assay explicitly:

Seurat::DefaultAssay(seu) <- "SCT"

SCT Note

If an object contains multiple SCT models, run PrepSCTFindMarkers() before marker testing.

This is common after merging SCT-normalized objects.

seu <- Seurat::PrepSCTFindMarkers(seu)

This prepares corrected counts for marker detection across SCT models.

Cluster Markers

Find markers for all clusters:

markers_all <- Seurat::FindAllMarkers(
  object = seu,
  assay = "SCT",
  only.pos = TRUE,
  min.pct = 0.25,
  logfc.threshold = 0.25,
  test.use = "wilcox",
  verbose = TRUE
)

Common parameters:

Parameter	Meaning
only.pos	return only genes enriched in the target group
min.pct	minimum fraction of cells expressing the gene in either group
logfc.threshold	minimum log fold change threshold
test.use	statistical test

The result is a data frame of marker genes.

Common result columns:

Column	Meaning
p_val	raw p-value before multiple-testing correction
avg_log2FC	average log2 fold change; positive values mean higher expression in the target group
pct.1	fraction of cells expressing the gene in group 1
pct.2	fraction of cells expressing the gene in group 2
p_val_adj	adjusted p-value, usually Bonferroni-corrected in Seurat
cluster	cluster associated with the marker; returned by FindAllMarkers()
gene	gene name

Check top markers:

head(markers_all)

Get top markers per cluster:

top_markers <- markers_all |>
  dplyr::group_by(cluster) |>
  dplyr::slice_max(
    order_by = avg_log2FC,
    n = 10,
    with_ties = FALSE
  ) |>
  dplyr::ungroup()

top_markers

One Cluster Versus Others

Find markers for one cluster:

cluster_0_markers <- Seurat::FindMarkers(
  object = seu,
  assay = "SCT",
  ident.1 = "0",
  only.pos = TRUE,
  min.pct = 0.25,
  logfc.threshold = 0.25,
  test.use = "wilcox",
  verbose = TRUE
)

This compares cluster 0 against all other cells by default.

Cluster Versus Cluster

Compare two clusters directly:

cluster_0_vs_1 <- Seurat::FindMarkers(
  object = seu,
  assay = "SCT",
  ident.1 = "0",
  ident.2 = "1",
  min.pct = 0.25,
  logfc.threshold = 0.25,
  test.use = "wilcox",
  verbose = TRUE
)

This compares cluster 0 against cluster 1.

Conserved Markers

FindConservedMarkers() finds markers for one identity that are conserved across groups such as samples, batches, or conditions.

Typical question:

Which markers define cluster 3 consistently across control and treatment?

Example:

cluster_3_conserved <- Seurat::FindConservedMarkers(
  object = seu,
  assay = "SCT",
  ident.1 = "3",
  grouping.var = "condition",
  test.use = "wilcox",
  only.pos = TRUE,
  min.pct = 0.25,
  logfc.threshold = 0.25,
  verbose = TRUE
)

This tests cluster 3 against other cells within each level of condition, then combines the evidence across groups.

Important parameters:

Parameter	Meaning
ident.1	target cluster or identity
grouping.var	metadata column used to split groups for conservation testing
only.pos	return only genes enriched in the target identity
min.pct	minimum fraction of cells expressing the gene in either group
logfc.threshold	minimum log fold change threshold
test.use	statistical test

Use conserved markers when annotating clusters across multiple samples or conditions. Do not use this as a replacement for condition differential expression.

Function	Question
FindAllMarkers()	What marks each cluster?
FindMarkers()	What differs between two selected groups?
FindConservedMarkers()	What marks one cluster consistently across groups?

Condition Markers

For condition comparison, compare conditions within a cell type or cluster.

First subset the relevant cells:

t_cells <- subset(
  x = seu,
  subset = cell_type == "T cell"
)

Then compare conditions:

condition_markers <- Seurat::FindMarkers(
  object = t_cells,
  assay = "SCT",
  ident.1 = "treated",
  ident.2 = "control",
  group.by = "condition",
  min.pct = 0.25,
  logfc.threshold = 0.25,
  test.use = "wilcox",
  verbose = TRUE
)

This asks which genes differ between treated and control cells within T cells.

Filter Results

Filter marker results:

markers_filtered <- markers_all |>
  dplyr::filter(
    p_val_adj < 0.05,
    avg_log2FC > 0.25
  )

Visualize Markers

Feature plot:

Seurat::FeaturePlot(
  object = seu,
  features = c("MS4A1", "CD3D"),
  reduction = "umap",
  pt.size = 0.5,
  alpha = 0.7,
  label = TRUE, # show cluster labels
  repel = TRUE # avoid label overlap
)

Violin plot:

Seurat::VlnPlot(
  object = seu,
  features = c("MS4A1", "CD3D"),
  group.by = "seurat_clusters",
  pt.size = 0
)

Dot plot:

Seurat::DotPlot(
  object = seu,
  features = c("MS4A1", "CD3D", "LYZ"),
  group.by = "seurat_clusters",
  dot.scale = 6
) +
  ggplot2::coord_flip()

Note

Marker results depend on the assay, identities, grouping variable, statistical test, filtering thresholds, and whether the comparison is biologically appropriate.

Cluster markers help interpret clusters. Condition markers answer differential-expression questions within a defined biological group.