I have a dataset from non-genomic data. 150 compounds was measured in leaves after a insect contact and non-insect contact (control). I have leaves in resistant and susceptible genotype. A subset of my data can be found at here.

Supposing Compound (Composto vary from 1 to 150) are genes, I used limma to get "differential" compunds between Genotypes, Estado (Resistante or Suscptible) and Tratamento (SI = without insect; CI = with insect).

I suppose that insect can increase or decrease some compounds in leaves and I suppose that resistant genotypes differ from susceptible ones in compounds productions.

My doubts:

1. I could use limma in this type of data?
2. I see that some compounds values are not normal distributed. I think to use box-cox to transform it. There are a way to found the lambda for each non normal row and do the conversion on the fly in MAList object only to that row (compound)?
3. If not possible to use box-cox, I think to use vsn or quantile to normalize the data. What they you advice? Or another one?

The targets are here.

I used this code to do an exploratory analysis:

library(limma)

dados01 <- read.table("Exemplo.csv", sep="\t", dec=",", header=TRUE, row.names = 1)

head(dados01)

targets <- readTargets("Targets.csv")
targets

### Tratamento

Group <- factor(targets$Tratamento, levels=c("SI","CI","PSI","PCI"))

design <- model.matrix(~0+Group)
colnames(design) <- c("SI", "CI","PSI","PCI")
design

contrast.matrix <- makeContrasts(CI-SI, CI-PCI,CI-PSI,SI-PCI,SI-PSI, levels=design)

contrast.matrix

E <- new("EList")

E$E <-dados01

E$genes <-rownames(dados01)

E2 <- normalizeBetweenArrays(E)

E2 <- normalizeVSN(E)

fit <- lmFit(E2, design)

fit2 <- contrasts.fit(fit, contrast.matrix)

fit2 <- eBayes(fit2)

topTable(fit2, coef=1, adjust="BH")

results <- decideTests(fit2)

vennDiagram(results)

results

### Estado

Group <- factor(targets$Estado, levels=c("Resistente","Suscetivel","Padrao"))

design <- model.matrix(~0+Group)

colnames(design) <- c("Resistente", "Suscetivel", "Padrao")

design

contrast.matrix <- makeContrasts(Resistente-Suscetivel, Resistente-Padrao, Suscetivel-Padrao, levels=design)

contrast.matrix

fit <- lmFit(E2, design)

fit2 <- contrasts.fit(fit, contrast.matrix)

fit2 <- eBayes(fit2)

topTable(fit2, coef=1, adjust="BH")

results <- decideTests(fit2)

vennDiagram(results)

results

Any comments here are very welcome!