Skip to contents

Perform Gene Set Variation Analysis (GSVA)

Source: R/RunGSVA.R
RunGSVA.Rd

Perform Gene Set Variation Analysis (GSVA)

Usage

RunGSVA(
  srt = 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,
  verbose = TRUE,
  ...
)

Arguments

srt

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

Which assay to use. If NULL, the default assay of the Seurat object will be used. When the object also contains ChromatinAssay, the default assay and additional ChromatinAssay will be preprocessed sequentially.

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. Default is "data".

assay_name

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

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

A character vector specifying the annotation sources to be included in the gene annotation databases. Can be one or more of "GO", "GO_BP", "GO_CC", "GO_MF", "KEGG", "WikiPathway", "Reactome", "CORUM", "MP", "DO", "HPO", "PFAM", "CSPA", "Surfaceome", "SPRomeDB", "VerSeDa", "TFLink", "hTFtarget", "TRRUST", "JASPAR", "ENCODE", "MSigDB", "CellTalk", "CellChat", "Chromosome", "GeneType", "Enzyme", "TF", "CytoTRACE2". MSigDB subcollections can be requested as "MSigDB_<collection>", such as "MSigDB_H" for human Hallmark and "MSigDB_MH" for mouse Hallmark. Note: "CytoTRACE2" is species-independent and downloads pre-trained model data required by RunCytoTRACE.

species

A character vector specifying the species for which the gene annotation databases should be prepared. Can be "Homo_sapiens" or "Mus_musculus".

IDtype

A character vector specifying the type of gene IDs in the srt 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

Whether the gene annotation databases should be forcefully updated. If set to FALSE, the function will attempt to load the cached databases instead. Default is FALSE.

db_version

A character vector specifying the version of the gene annotation databases to be retrieved. Default is "latest".

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

Whether to use a species-converted database when the annotation is missing for the specified species. Default is TRUE.

Ensembl_version

An integer specifying the Ensembl version. Default is NULL. If NULL, the latest version will be used.

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

A character vector specifying the 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". Default is "gsva".

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". 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

Logical indicating whether to use absolute ranking for GSVA. Default is FALSE.

min.sz

Minimum size of gene sets to be included in the analysis. Default is 10.

max.sz

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

mx.diff

Logical indicating whether to use the maximum difference method. Default is TRUE.

tau

Exponent for the GSVA method. Default is 1.

ssgsea.norm

Logical indicating whether to normalize SSGSEA scores. Default is TRUE.

verbose

Whether to print the message. Default is TRUE.

...

Passed to other functions.

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 <- standard_scop(pancreas_sub)
#> ℹ [2026-07-02 09:42:18] Start standard processing workflow...
#> ℹ [2026-07-02 09:42:19] Checking a list of <Seurat>...
#> ! [2026-07-02 09:42:19] Data 1/1 of the `srt_list` is "unknown"
#> ℹ [2026-07-02 09:42:19] Perform `NormalizeData()` with `normalization.method = 'LogNormalize'` on 1/1 of `srt_list`...
#> ℹ [2026-07-02 09:42:19] Perform `FindVariableFeatures()` on 1/1 of `srt_list`...
#> ℹ [2026-07-02 09:42:19] Use the separate HVF from `srt_list`
#> ℹ [2026-07-02 09:42:19] Number of available HVF: 2000
#> ℹ [2026-07-02 09:42:19] Finished check
#> ℹ [2026-07-02 09:42:19] Perform `ScaleData()`
#> ℹ [2026-07-02 09:42:19] Perform pca linear dimension reduction
#> ℹ [2026-07-02 09:42:20] Use stored estimated dimensions 1:23 for Standardpca
#> ℹ [2026-07-02 09:42:20] Perform `Seurat::FindClusters()` with `cluster_algorithm = 'louvain'` and `cluster_resolution = 0.6`
#> ℹ [2026-07-02 09:42:20] Reorder clusters...
#> ℹ [2026-07-02 09:42:20] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-07-02 09:42:20] Perform umap nonlinear dimension reduction
#> ✔ [2026-07-02 09:42:27] Standard processing workflow completed

pancreas_sub <- RunGSVA(
  pancreas_sub,
  group.by = "CellType",
  species = "Mus_musculus"
)
#> ℹ [2026-07-02 09:42:27] Start GSVA analysis
#> ℹ [2026-07-02 09:42:27] Start GSVA analysis
#> ℹ [2026-07-02 09:42:27] Species: "Mus_musculus"
#> ℹ [2026-07-02 09:42:27] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-07-02 08:59:44
#> ℹ [2026-07-02 09:42:28] Averaging expression by "CellType" ...
#> ℹ [2026-07-02 09:42:29] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-07-02 09:42:29] Processing database: "GO_BP" ...
#> ℹ [2026-07-02 09:42:30] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-07-02 09:42:30] Running GSVA for 5633 gene sets ...
#> ℹ 47266 nonzeros (less than 2^31) and 16.17% sparsity
#> ℹ [2026-07-02 09:42:34] GSVA results stored in `tools` slot: "GSVA_CellType_gsva"
#> ✔ [2026-07-02 09:42:34] GSVA analysis done
#> ℹ [2026-07-02 09:42:34] Start GSVA analysis
#> ℹ [2026-07-02 09:42:34] Species: "Mus_musculus"
#> ℹ [2026-07-02 09:42:34] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-07-02 08:59:44
#> ℹ [2026-07-02 09:42:35] Averaging expression by "CellType" ...
#> ℹ [2026-07-02 09:42:35] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-07-02 09:42:35] Processing database: "GO_BP" ...
#> ℹ [2026-07-02 09:42:37] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-07-02 09:42:37] Running GSVA for 5633 gene sets ...
#> ℹ [2026-07-02 09:42:38] GSVA results stored in `tools` slot: "GSVA_CellType_ssgsea"
#> ✔ [2026-07-02 09:42:38] GSVA analysis done
#> ℹ [2026-07-02 09:42:38] Start GSVA analysis
#> ℹ [2026-07-02 09:42:38] Species: "Mus_musculus"
#> ℹ [2026-07-02 09:42:38] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-07-02 08:59:44
#> ℹ [2026-07-02 09:42:40] Averaging expression by "CellType" ...
#> ℹ [2026-07-02 09:42:40] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-07-02 09:42:40] Processing database: "GO_BP" ...
#> ℹ [2026-07-02 09:42:41] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-07-02 09:42:41] Running GSVA for 5633 gene sets ...
#> ℹ [2026-07-02 09:42:43] GSVA results stored in `tools` slot: "GSVA_CellType_zscore"
#> ✔ [2026-07-02 09:42:43] GSVA analysis done
#> ℹ [2026-07-02 09:42:43] Start GSVA analysis
#> ℹ [2026-07-02 09:42:43] Species: "Mus_musculus"
#> ℹ [2026-07-02 09:42:43] Loading cached: GO_BP version: 3.23.0 nterm:14957 created: 2026-07-02 08:59:44
#> ℹ [2026-07-02 09:42:44] Averaging expression by "CellType" ...
#> ℹ [2026-07-02 09:42:44] Aggregated expression matrix: 15998 genes x 5 groups
#> ℹ [2026-07-02 09:42:44] Processing database: "GO_BP" ...
#> ℹ [2026-07-02 09:42:45] Initial overlap: 11277 genes out of 15998 expression genes and 16594 genes in gene sets
#> ℹ [2026-07-02 09:42:45] Running GSVA for 5633 gene sets ...
#> ℹ [2026-07-02 09:42:51] GSVA results stored in `tools` slot: "GSVA_CellType_plage"
#> ✔ [2026-07-02 09:42:51] GSVA analysis done
ht <- GSVAPlot(
  pancreas_sub,
  group.by = "CellType",
  plot_type = "heatmap",
  topTerm = 10,
  width = 1,
  height = 2
)
#> ! [2026-07-02 09:42:51] Multiple GSVA results found for "CellType". Using "GSVA_CellType_gsva"
#> Warning: Data is of class matrix. Coercing to dgCMatrix.

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 [2026-07-02 09:42:52] Start GSVA analysis
#> ℹ [2026-07-02 09:42:52] Start GSVA analysis
#> ℹ [2026-07-02 09:42:52] Single-cell GSVA mode: using expression matrix directly ...
#> ℹ [2026-07-02 09:42:52] Expression matrix: 15998 genes x 1000 cells
#> ℹ [2026-07-02 09:42:52] Processing database: "custom" ...
#> ℹ [2026-07-02 09:42:52] Initial overlap: 40 genes out of 15998 expression genes and 40 genes in gene sets
#> ℹ [2026-07-02 09:42:52] Running GSVA for 2 gene sets ...
#> ℹ 6830 nonzeros (less than 2^31) and 82.92% sparsity
#> Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
#> Warning: Layer counts isn't present in the assay object; returning NULL
#> ℹ [2026-07-02 09:42:52] GSVA results stored in assay "GSVA_gsva", meta.data, and tools slot "GSVA_cell_gsva"
#> ✔ [2026-07-02 09:42:52] GSVA analysis done
#> ℹ [2026-07-02 09:42:52] Start GSVA analysis
#> ℹ [2026-07-02 09:42:52] Single-cell GSVA mode: using expression matrix directly ...
#> ℹ [2026-07-02 09:42:52] Expression matrix: 15998 genes x 1000 cells
#> ℹ [2026-07-02 09:42:52] Processing database: "custom" ...
#> ℹ [2026-07-02 09:42:52] Initial overlap: 40 genes out of 15998 expression genes and 40 genes in gene sets
#> ℹ [2026-07-02 09:42:52] Running GSVA for 2 gene sets ...
#> Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
#> Warning: Layer counts isn't present in the assay object; returning NULL
#> ℹ [2026-07-02 09:42:52] GSVA results stored in assay "GSVA_ssgsea", meta.data, and tools slot "GSVA_cell_ssgsea"
#> ✔ [2026-07-02 09:42:52] GSVA analysis done
#> Warning: Key ‘gsvagsva_’ taken, using ‘gsva_’ instead
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 [2026-07-02 09:42:52] Setting `group.by` to "Features" as `plot.by` is set to "feature"