GSVA Algorithm

Purpose and Usage

• Scenario 1: Perform GSVA analysis on specific objects, e.g., specify clusters 1, 2, and 3 in leiden clustering for GSVA analysis

  SDAS geneSetEnrichment gsva -i st.h5ad -o outdir \
  --group_key leiden --idents 1,2,3 --species human

• Scenario 2: Subset a column in obs before GSVA analysis, e.g., perform GSVA analysis only on different samples of a certain cell type

  SDAS geneSetEnrichment gsva -i st.h5ad -o outdir \
  --group_key leiden --idents 1,2,3 --species human \
  --subset_key cell_type --subset_values B

• Scenario 3: Analyze only with databases of interest

  SDAS geneSetEnrichment gsva -i st.h5ad -o outdir \
  --group_key leiden --idents 1,2,3 \
  --gmt sdas_deg_enrichment/lib/GSEADB/h.all.v2024.1.Hs.symbols.gmt,sdas_deg_enrichment/lib/GSEADB/KEGG_2021_Human.gmt

• Scenario 4: Perform GSVA analysis on all elements of a column in obs together. To analyze all elements in a column together, set the --idents parameter to the special value all.

  SDAS geneSetEnrichment gsva -i st.h5ad -o outdir \
  --group_key leiden --idents all --species human \

Input Parameter Description

GSVA Parameter	Required	Default Value	Description
-i / --input	Yes		Input a spatial h5ad file.
--group_key	Yes		Identifier name in h5ad obs, must contain ident1 and ident2.
--idents	Yes		Regroup cells into a different identity class prior to performing GSVA, eg. sample1,sample2, default: all
-o / --output	Yes		The GSEApy output directory. Default: the current working directory
--subset_key	No		Key for subsetting (optional), eg. cell_type
--subset_values	No		Values used for subsetting (optional), eg. cell1,cell2
--layer	No		set gene raw expression layer, adata.raw.X or adata.X will be used if set None . default: None
--gene_symbol_key	No	real_gene_name	set gene name, default: real_gene_name
--species	No	human	Use biuld-in gmt database: human or mouse. Default: human. More database see here: https://amp.pharm.mssm.edu/modEnrichr.
--sample_size	No	0	Random sample cells number, 0 for not. Default: 0
--gmt	No		Customized gene set database in GMT format. One or more databases split by ",". Default use --species build-in database.
--kernel_cdf	No	Gaussian	Gaussian is suitable when input expression values are continuous. If input integer counts, then this argument should be set to 'Poisson'
--mx_diff	No		When set, ES is calculated as the maximum distance of the random walk from 0. Default: False
--abs_ranking	No		Flag used only when --mx-diff is not set. When set, the original Kuiper statistic is used
--min_size	No	15	Min size of input genes presented in Gene Sets. Default: 15
--max_size	No	20000	Max size of input genes presented in Gene Sets. Default: 20000
--weight	No	1	tau in the random walk performed by the gsva. Default: 1
--seed	No	123	Number of random seed. Default: 123
--threads	No	1	Number of Processes you are going to use. Default: 1

Output Results Display

GSVA Result File	Description
GSVA.{database}.csv	Result file in csv format
GSVA.{database}.pdf/png	Result heatmap in pdf and png formats

• GSVA csv file format: GSVA.{database}.csv is the GSVA analysis result file. The first column is Term (function name), and each subsequent column represents a sample. Positive values indicate higher activity for that function in the sample, negative values indicate lower activity.

Term	ident1	ident2	...
HALLMARK_ADIPOGENESIS	-0.32809425650271146	-0.306805475112318	....
HALLMARK_ALLOGRAFT_REJECTION	-0.3052190348950549	0.22055475913564931	....
HALLMARK_ANDROGEN_RESPONSE	-0.39290236695613107	-0.3080397441881526	....
...	...	...	...

• GSVA result heatmap: GSVA.{database}.pdf/png. The vertical axis is the function/pathway name, the horizontal axis is the sample name, and the legend shows the GSVA score.