library(devtools)
install_github("guokai8/scGSVA")

# For UCell support (optional)
BiocManager::install("UCell")

set.seed(123)
library(scGSVA)
data(pbmcs)

# Build annotation
hsko <- buildAnnot(species = "human", keytype = "SYMBOL", anntype = "KEGG")

# Run enrichment analysis (uses fast C++ implementation by default)
res <- scgsva(pbmcs, hsko, method = "ssgsea")

# Other methods: gsva, plage, zscore
res <- scgsva(pbmcs, hsko, method = "gsva")
res <- scgsva(pbmcs, hsko, method = "plage")
res <- scgsva(pbmcs, hsko, method = "zscore")

# Or use UCell (maxRank is auto-adjusted to fit gene set sizes)
res <- scgsva(pbmcs, hsko, method = "UCell")

vlnPlot(res, features = "Wnt.signaling.pathway", group_by = "groups")

# Split violin plot (compare conditions within each cluster)
vlnPlot(res, features = "Wnt.signaling.pathway", group_by = "seurat_clusters",
        split.by = "groups", split.plot = TRUE)

dotPlot(res, features = "Wnt.signaling.pathway", group_by = "groups")

ridgePlot(res, features = "Wnt.signaling.pathway", group_by = "groups")

featurePlot(res, features = "Wnt.signaling.pathway", reduction = "tsne")

Heatmap(res, group_by = "groups")

barPlot(res, features = c("Wnt.signaling.pathway", "MAPK.signaling.pathway"),
        group_by = "groups")

lollipopPlot(res, features = c("Wnt.signaling.pathway", "MAPK.signaling.pathway"),
             group_by = "groups")

# Linear model-based analysis (limma)
findPathway(res, group = "groups")

# Statistical tests (t-test, Wilcoxon, ANOVA)
sigPathway(res, group = "groups", test.use = "wilcox")

# Get all genes in a pathway with expression values
genes(res, features = "Wnt.signaling.pathway")

# Get top influential genes driving pathway scores
topGenes(res, features = "Wnt.signaling.pathway", n = 10)

# With group-wise statistics
topGenes(res, features = "Wnt.signaling.pathway", n = 10, group = "groups")

# Get summary statistics for pathways
summaryPathway(res, features = c("Wnt.signaling.pathway", "MAPK.signaling.pathway"),
               group_by = "groups")

# KEGG pathways
hsko <- buildAnnot(species = "human", keytype = "SYMBOL", anntype = "KEGG")

# GO terms
hsgo <- buildAnnot(species = "human", keytype = "SYMBOL", anntype = "GO")

# Check supported species
showData()

# Hallmark gene sets
hallmark <- buildMSIGDB(species = "human", keytype = "SYMBOL",
                        anntype = "HALLMARK")

# KEGG pathways from MSigDB
msig_kegg <- buildMSIGDB(species = "human", keytype = "SYMBOL",
                         anntype = "KEGG")

# GO Biological Process
go_bp <- buildMSIGDB(species = "human", keytype = "SYMBOL",
                     anntype = "BP")

# Check available annotation types
msigdbinfo()

# 1. Download manually (use VPN or mirror):
#    Human: https://zenodo.org/records/15800824/files/msigdb.2025.1.Hs.rds
#    Mouse: https://zenodo.org/records/15800824/files/msigdb.2025.1.Mm.rds

# 2. Load and use:
msig_data <- readRDS("msigdb.2025.1.Hs.rds")
hallmark <- buildMSIGDB(species = "human", keytype = "SYMBOL",
                        anntype = "HALLMARK", msigdb_data = msig_data)

library(Seurat)
library(SeuratData)

# Load spatial data
brain <- LoadData("stxBrain", type = "anterior1")

# Run enrichment
hsko <- buildAnnot(species = "human", keytype = "SYMBOL", anntype = "KEGG")
res <- scgsva(brain, hsko, assay = "Spatial")

# Visualize on tissue
spatialFeaturePlot(res, features = "Wnt.signaling.pathway")

# Default: fast custom C++ implementation
res <- scgsva(pbmcs, hsko, method = "ssgsea")

# Original GSVA package (for comparison/validation)
res_orig <- scgsva(pbmcs, hsko, method = "ssgsea", use.original = TRUE)