**10**wrote:

Hello, I'm working on GSVA and further analysis using Limma just like the https://bioconductor.org/packages/release/bioc/vignettes/GSVA/inst/doc/GSVA.pdf figure 4.

My data is clustered into 3 types, according to the TCGA protocol for LUAD. I followed the description from the GSVA.pdf and the answers from biostars written by Kevin Blighe.

{My R code for analysis}

```
lung_es <- gsva(exampleSet, gm, method = "gsva", min.sz=10, max.sz=9999, verbose=TRUE, abs.ranking=FALSE, no.bootstraps=0, bootstrap.percent = .632)
adjPvalueCutoff <- 0.05 #P-value
logFCcutoff <- log2(2)
design <- model.matrix(~factor(lung_es$TCGA_subtype))
colnames(design) <- c("TRU_Bron", "PI_Squam", "PP_Magn")
fit <- lmFit(lung_es, design)
fit <- eBayes(fit)
allGeneSets <- topTable(fit, coef=NULL, number=Inf)
DEgeneSets <- topTable(fit, coef=NULL, number=Inf, p.value=adjPvalueCutoff, adjust="BH")
res <- decideTests(fit, p.value=adjPvalueCutoff)
summary(res)
# TRU_Bron PI_Squam PP_Magn
#Down 8 168 209
#NotSig 1143 834 793
#Up 13 162 162
sig.genes <- c(names(res[res[,2]==1,2]), names(res[res[,2]==-1,2]))
lung_eset.siggenes <- lung_es[which(rownames(lung_es) %in% sig.genes),]
```

so, this was the answer and I followed without question...
**turns out the example (leukemia) has two subtype (ALL, MLL)**

**BUT, I have three subtype. So i think that code getting sig.genes should be different.**

*Any suggestion about getting differently activated or deactivated pathways among three subtypes??*