Dear Community,

based on some preliminary in vitro data about a specific microRNA that is related with stemness, I analyzed a small number of relative microarray samples, based on two conditions (3 reps each)-one of the major downstream goals, was to identify if any pathways/processes related to the Notch signaling pathway, that is of high interest are found to be collectively DE as a gene set, which led me to use mroast:

# initial pre-processing-normalization-filtering
condition <- factor(eset.filtered$Condition, levels=c("Control","Transfected"))
design <- model.matrix(~condition)
fit <- lmFit(eset.filtered,design)
fit2 <- eBayes(fit, trend=TRUE,robust = TRUE)

# general pathways from msigdb related to NOTCH
library(qusage)
pathway.sets <- read.gmt("c2.cp.v7.2.symbols.gmt")
pathway.notch.sets <- pathway.sets[grep("NOTCH", names(pathway.sets))]
indices <- ids2indices(pathway.notch.sets, fData(eset.filtered)$SYMBOL)
set.size <- sapply(indices, FUN=length)
indices2 <- indices[set.size >= 10]

res <- mroast(eset.filtered, indices2, design=design, contrast=2,nrot=10000)

head(res)
                                                                    NGenes   PropDown    PropUp
REACTOME_NOTCH3_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION            57 0.10526316 0.1929825
WP_NOTCH1_REGULATION_OF_HUMAN_ENDOTHELIAL_CELL_CALCIFICATION            44 0.22727273 0.1136364
REACTOME_NOTCH4_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION            43 0.09302326 0.2790698
WP_NOTCH_SIGNALING_PATHWAY_NETPATH                                     167 0.13173653 0.2814371
WP_APOPTOSISRELATED_NETWORK_DUE_TO_ALTERED_NOTCH3_IN_OVARIAN_CANCER    157 0.26114650 0.1656051
REACTOME_PRE_NOTCH_EXPRESSION_AND_PROCESSING                           151 0.29801325 0.1456954
                                                                    Direction     PValue        FDR
REACTOME_NOTCH3_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION               Up 0.00019998 0.00479952
WP_NOTCH1_REGULATION_OF_HUMAN_ENDOTHELIAL_CELL_CALCIFICATION             Down 0.00069993 0.01039896
REACTOME_NOTCH4_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION               Up 0.00419958 0.03862471
WP_NOTCH_SIGNALING_PATHWAY_NETPATH                                         Up 0.00709929 0.03862471
WP_APOPTOSISRELATED_NETWORK_DUE_TO_ALTERED_NOTCH3_IN_OVARIAN_CANCER      Down 0.00769923 0.03862471
REACTOME_PRE_NOTCH_EXPRESSION_AND_PROCESSING                             Down 0.00809919 0.03862471
                                                                    PValue.Mixed   FDR.Mixed
REACTOME_NOTCH3_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION          0.01819818 0.022336228
WP_NOTCH1_REGULATION_OF_HUMAN_ENDOTHELIAL_CELL_CALCIFICATION          0.00029997 0.000799920
REACTOME_NOTCH4_INTRACELLULAR_DOMAIN_REGULATES_TRANSCRIPTION          0.00239976 0.003269238
WP_NOTCH_SIGNALING_PATHWAY_NETPATH                                    0.00019998 0.000799920
WP_APOPTOSISRELATED_NETWORK_DUE_TO_ALTERED_NOTCH3_IN_OVARIAN_CANCER   0.00209979 0.002981520
REACTOME_PRE_NOTCH_EXPRESSION_AND_PROCESSING                          0.00079992 0.001333200

My main question is, for the interpretation of the results, and to emphasize that indeed most of the NOTCH-related pathways are found to be DE by mroast, should I focus on mentioning FDR.Mixed ? as FDR.Mixed is more suitable when the genes in the gene-set are non perhaps co-regulated, or even in different directions ? and the focus is primarily if the vast majority of the selected pathways are found DE ?

Alternatively, camera gene-set testing could be applied for accounting also for inter-gene correlations, and a more direct approach ?

Best,

Efstathios