Skip to contents

Perform Gene Set Variation Analysis (GSVA)

Source: R/RunGSVA.R
RunGSVA.Rd

Perform Gene Set Variation Analysis (GSVA)

Usage

RunGSVA(
  object = NULL,
  assay = NULL,
  group.by = NULL,
  layer = "data",
  assay_name = "GSVA",
  new_assay = TRUE,
  store_metadata = NULL,
  db = "GO_BP",
  species = "Homo_sapiens",
  IDtype = "symbol",
  db_update = FALSE,
  db_version = "latest",
  db_combine = FALSE,
  convert_species = TRUE,
  Ensembl_version = NULL,
  mirror = NULL,
  features = NULL,
  TERM2GENE = NULL,
  TERM2NAME = NULL,
  minGSSize = 10,
  maxGSSize = 500,
  unlimited_db = c("Chromosome", "GeneType", "TF", "Enzyme", "CSPA"),
  method = c("gsva", "ssgsea", "zscore", "plage"),
  backend = c("cpp", "r"),
  cpp_chunk_size = NULL,
  kcdf = c("Gaussian", "Poisson", "none"),
  abs.ranking = FALSE,
  min.sz = 10,
  max.sz = Inf,
  mx.diff = TRUE,
  tau = 1,
  ssgsea.norm = TRUE,
  cores = NULL,
  verbose = TRUE,
  ...,
  srt = NULL
)

Arguments

object

A Seurat object or SummarizedExperiment object containing the results of differential expression analysis (RunDEtest()). If specified, the genes and groups will be extracted from the object automatically. If not specified, the geneID and geneID_groups arguments must be provided.

assay

Assay to use. NULL uses the default assay.

group.by

Name of metadata column to group cells by for averaging expression. If provided, expression will be averaged within each group before GSVA analysis (cell-type level). If NULL, GSVA is performed on each cell individually (single-cell level).

layer

Data layer to use when group.by = NULL. Usually "data" for normalized or "counts" for count matrix.

assay_name

Name of the assay to store GSVA scores when group.by = NULL and new_assay = TRUE.

new_assay

Whether to create a new assay for GSVA scores when group.by = NULL. Default is TRUE.

store_metadata

Whether to also store single-cell GSVA scores in meta.data. When NULL, custom features or TERM2GENE input is stored in meta.data by default, while database-derived results stay assay-only when new_assay = TRUE.

db

Character vector of database or resource selectors: "GO", "GO_BP", "GO_CC", "GO_MF", "KEGG", "WikiPathway", "Reactome", "CORUM", "MP", "DO", "HPO", "PFAM", "Chromosome", "GeneType", "Enzyme", "TF", "CSPA", "Surfaceome", "SPRomeDB", "VerSeDa", "TFLink", "hTFtarget", "TRRUST", "JASPAR", "ENCODE", "MSigDB", "CellTalk", "CellChat", "CytoTRACE2" or "MSigDB_<collection>". A vector may combine sources. See the corresponding preparation function for source-specific settings.

species

"Homo_sapiens" or "Mus_musculus".

IDtype

Type of gene IDs in the object object or geneID argument. This argument is used to convert the gene IDs to a different type if IDtype is different from result_IDtype.

db_update

Force a refresh. FALSE loads the cache when available.

db_version

Database version to retrieve.

db_combine

Whether to combine multiple databases into one. If TRUE, all database specified by db will be combined as one named "Combined".

convert_species

Use a species-converted database when the annotation is missing for species.

Ensembl_version

Ensembl version. NULL uses the latest.

mirror

Specify an Ensembl mirror to connect to. The valid options here are "www", "uswest", "useast", "asia".

features

A named list of feature lists for custom enrichment gene sets. If provided, it takes precedence over TERM2GENE and db.

TERM2GENE

A data frame specifying the gene-term mapping for a custom database. The first column should contain the term IDs, and the second column should contain the gene IDs.

TERM2NAME

A data frame specifying the term-name mapping for a custom database. The first column should contain the term IDs, and the second column should contain the corresponding term names.

minGSSize

The minimum size of a gene set to be considered in the enrichment analysis.

maxGSSize

The maximum size of a gene set to be considered in the enrichment analysis.

unlimited_db

Names of databases that do not have size restrictions.

method

The method to use for GSVA. Options are "gsva", "ssgsea", "zscore", or "plage". Multiple methods can be supplied at once; in single-cell mode they will be stored in method-suffixed assays such as "GSVA_gsva" and "GSVA_ssgsea".

backend

Scoring backend. "cpp" is the default and supports all current method values. "r" uses the original GSVA::gsva() implementation. "cpp" supports method = "ssgsea", method = "zscore", method = "plage", and method = "gsva" with kcdf = "Gaussian", kcdf = "Poisson", or kcdf = "none". Gaussian GSVA uses the native C++ KDE/ranking kernel, tracking the sparse scoring walk of the installed GSVA release; other GSVA kernels keep the validated GSVA implementation. PLAGE scores are oriented to have non-negative dot product with the gene set mean z-score so SVD signs are deterministic.

cpp_chunk_size

Optional cell chunk size for C++ GSVA kernels. NULL or "auto" automatically chunks large matrices to reduce peak dense intermediate memory; positive values set the chunk size manually.

kcdf

The kernel cumulative distribution function used for GSVA. Options are "Gaussian" (for continuous data), "Poisson" (for count data), or "none" (skip kernel estimation and use ranks directly). When omitted, backend = "cpp" with method = "gsva" uses "none" for faster single-cell scoring; explicit "Gaussian" or "Poisson" values are still honored. Other backends and methods default to "Gaussian".

abs.ranking

Whether to use absolute ranking for GSVA.

min.sz

Minimum size of gene sets to be included in the analysis.

max.sz

Maximum size of gene sets to be included in the analysis.

mx.diff

Whether to use the maximum difference method.

tau

Exponent for the GSVA method.

ssgsea.norm

Whether to normalize SSGSEA scores.

cores

Number of OpenMP threads for C++ scoring. NULL uses the process OpenMP default.

verbose

Whether to print the message. Default is TRUE.

...

Passed to helper functions.

srt

Deprecated alias for object; supply exactly one of the two. It will be removed in scop 1.0.0.

Value

Returns the modified Seurat object. When group.by is provided, GSVA scores are stored in the tools slot. When group.by = NULL, scores are stored in the tools slot, optionally in a new assay, and optionally in meta.data for direct use with FeatureDimPlot() and FeatureStatPlot().

Examples

data(pancreas_sub)
pancreas_sub <- RunStandardWorkflow(pancreas_sub)
#> ℹ [2026-10-11 23:19:21] Start standard processing workflow...
#> ℹ [2026-10-11 23:19:21] Checking a list of <Seurat>...
#> ! [2026-10-11 23:19:21] Data 1/1 of the `srt_list` is "unknown"
#> ℹ [2026-10-11 23:19:21] Perform `NormalizeData()` with `normalization.method = 'LogNormalize'` on 1/1 of `srt_list`...
#> ℹ [2026-10-11 23:19:22] Perform `FindVariableFeatures()` on 1/1 of `srt_list`...
#> ℹ [2026-10-11 23:19:22] Use the separate HVF from `srt_list`
#> ℹ [2026-10-11 23:19:22] Number of available HVF: 2000
#> ℹ [2026-10-11 23:19:22] Finished check
#> ℹ [2026-10-11 23:19:22] Perform `ScaleData()`
#> ℹ [2026-10-11 23:19:22] Perform pca linear dimension reduction
#> ℹ [2026-10-11 23:19:22] Use stored estimated dimensions 1:23 for Standardpca
#> ℹ [2026-10-11 23:19:23] Perform `Seurat::FindClusters()` with `cluster_algorithm = 'louvain'` and `cluster_resolution = 0.6`
#> ℹ [2026-10-11 23:19:23] Reorder clusters...
#> ℹ [2026-10-11 23:19:23] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-10-11 23:19:23] Perform umap nonlinear dimension reduction
#> ✔ [2026-10-11 23:19:30] Standard processing workflow completed

pancreas_sub <- RunGSVA(
  pancreas_sub,
  group.by = "CellType",
  species = "Mus_musculus"
)
#> ℹ [2026-10-11 23:19:30] Start GSVA analysis
#> ℹ [2026-10-11 23:19:30] Start GSVA analysis
#> ℹ [2026-10-11 23:19:30] Species: "Mus_musculus"
#> ℹ [2026-10-11 23:19:30] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-10-11 22:48:45
#> ℹ [2026-10-11 23:19:31] Averaging expression by "CellType" ...
#> ℹ [2026-10-11 23:19:31] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-10-11 23:19:31] Processing database: "GO_BP" ...
#> ℹ [2026-10-11 23:19:32] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-10-11 23:19:32] Running GSVA for 5633 gene sets ...
#> ℹ 45447 nonzeros (less than 2^31) and 7.96% sparsity
#> ℹ [2026-10-11 23:19:38] GSVA results stored in `tools` slot: "GSVA_CellType_gsva"
#> ✔ [2026-10-11 23:19:38] GSVA analysis done
#> ℹ [2026-10-11 23:19:38] Start GSVA analysis
#> ℹ [2026-10-11 23:19:38] Species: "Mus_musculus"
#> ℹ [2026-10-11 23:19:38] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-10-11 22:48:45
#> ℹ [2026-10-11 23:19:39] Averaging expression by "CellType" ...
#> ℹ [2026-10-11 23:19:39] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-10-11 23:19:39] Processing database: "GO_BP" ...
#> ℹ [2026-10-11 23:19:40] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-10-11 23:19:40] Running GSVA for 5633 gene sets ...
#> ℹ [2026-10-11 23:19:42] GSVA results stored in `tools` slot: "GSVA_CellType_ssgsea"
#> ✔ [2026-10-11 23:19:42] GSVA analysis done
#> ℹ [2026-10-11 23:19:42] Start GSVA analysis
#> ℹ [2026-10-11 23:19:42] Species: "Mus_musculus"
#> ℹ [2026-10-11 23:19:42] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-10-11 22:48:45
#> ℹ [2026-10-11 23:19:43] Averaging expression by "CellType" ...
#> ℹ [2026-10-11 23:19:43] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-10-11 23:19:43] Processing database: "GO_BP" ...
#> ℹ [2026-10-11 23:19:44] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-10-11 23:19:44] Running GSVA for 5633 gene sets ...
#> ℹ [2026-10-11 23:19:46] GSVA results stored in `tools` slot: "GSVA_CellType_zscore"
#> ✔ [2026-10-11 23:19:46] GSVA analysis done
#> ℹ [2026-10-11 23:19:46] Start GSVA analysis
#> ℹ [2026-10-11 23:19:46] Species: "Mus_musculus"
#> ℹ [2026-10-11 23:19:46] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-10-11 22:48:45
#> ℹ [2026-10-11 23:19:46] Averaging expression by "CellType" ...
#> ℹ [2026-10-11 23:19:46] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-10-11 23:19:46] Processing database: "GO_BP" ...
#> ℹ [2026-10-11 23:19:48] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-10-11 23:19:48] Running GSVA for 5633 gene sets ...
#> ℹ [2026-10-11 23:19:49] GSVA results stored in `tools` slot: "GSVA_CellType_plage"
#> ✔ [2026-10-11 23:19:49] GSVA analysis done
ht <- GSVAPlot(
  pancreas_sub,
  group.by = "CellType",
  plot_type = "heatmap",
  topTerm = 10,
  width = 1,
  height = 2
)
#> ! [2026-10-11 23:19:49] Multiple GSVA results found for "CellType". Using "GSVA_CellType_gsva"

features_all <- rownames(pancreas_sub)
pancreas_sub <- RunGSVA(
  pancreas_sub,
  features = list(
    A = features_all[1:20],
    B = features_all[21:40]
  ),
  method = c("gsva", "ssgsea")
)
#> ℹ [2026-10-11 23:19:50] Start GSVA analysis
#> ℹ [2026-10-11 23:19:50] Start GSVA analysis
#> ℹ [2026-10-11 23:19:50] Single-cell GSVA mode: using expression matrix directly ...
#> ℹ [2026-10-11 23:19:50] Expression matrix: 15998 genes x 1000 cells
#> ℹ [2026-10-11 23:19:50] Processing database: "custom" ...
#> ℹ [2026-10-11 23:19:50] Initial overlap: 40 genes out of 15998 expression genes and 40 genes in gene sets
#> ℹ [2026-10-11 23:19:50] Running GSVA for 2 gene sets ...
#> ℹ 6826 nonzeros (less than 2^31) and 81.04% sparsity
#> ℹ [2026-10-11 23:19:52] GSVA results stored in assay "GSVA_gsva", meta.data, and tools slot "GSVA_cell_gsva"
#> ✔ [2026-10-11 23:19:52] GSVA analysis done
#> ℹ [2026-10-11 23:19:52] Start GSVA analysis
#> ℹ [2026-10-11 23:19:52] Single-cell GSVA mode: using expression matrix directly ...
#> ℹ [2026-10-11 23:19:52] Expression matrix: 15998 genes x 1000 cells
#> ℹ [2026-10-11 23:19:52] Processing database: "custom" ...
#> ℹ [2026-10-11 23:19:52] Initial overlap: 40 genes out of 15998 expression genes and 40 genes in gene sets
#> ℹ [2026-10-11 23:19:52] Running GSVA for 2 gene sets ...
#> ℹ [2026-10-11 23:19:52] GSVA results stored in assay "GSVA_ssgsea", meta.data, and tools slot "GSVA_cell_ssgsea"
#> ✔ [2026-10-11 23:19:52] GSVA analysis done
FeatureDimPlot(
  pancreas_sub,
  features = "GSVA_gsva_A",
  add_density = TRUE
)

FeatureStatPlot(
  pancreas_sub,
  stat.by = c("GSVA_gsva_A", "GSVA_ssgsea_A"),
  group.by = "CellType",
  plot.by = "feature",
  plot_type = "violin",
  stack = TRUE,
  flip = TRUE
)
#> ℹ [2026-10-11 23:19:52] Setting `group.by` to "Features" as `plot.by` is set to "feature"