Gordon, Any insight on the below? I am looking for a bit of enlightenment regarding edgeR. We are applying edgeR using ANOVA-like testing. We have 2 conditions (Cases and Controls) sampled at 2 treatments (untreated and treated) with 64 biological replicates per condition (64 X 2 = 128 samples). We want to test 3 ways: 1. DE between untreated and treated for cases. 2. DE between untreated and treated for controls. 3. DE between untreated and treated responding differently between Cases and Controls. I am curious if my design matrix is set up properly and if we are testing with the right coefficients and contrasts as seen below to address these three approaches. Any insight is helpful. Yours, Michael rawdata <- read.delim("Matrix.txt", check.names=FALSE, stringsAsFactors=FALSE) targets <- read.delim("Described.txt", check.names=FALSE, stringsAsFactors=FALSE) head(targets) Condition Patient Treatments 1 Case 4060 untreated 2 Case 4060 treated 3 Case 4096 untreated 4 Case 4096 treated 5 Case 4099 untreated 6 Case 4099 treated y <- DGEList(counts=rawdata[,2:129], genes=rawdata[,1:1]) keep <- rowSums(cpm(y)>1) >= 64 y <- y[keep,] dim (y) y <- calcNormFactors(y) y$samples design <- model.matrix(~Condition+Condition:Patient+Condition: Treatments, data=targets) colnames(design) [1] "(Intercept)" "ConditionControl" [3] "ConditionCase:Patient" "ConditionControl:Patient" [5] "ConditionCase:Treatments " "ConditionControl:Treatments " y <- estimateGLMCommonDisp(y, design, verbose=TRUE) y <- estimateGLMTrendedDisp(y, design) y <- estimateGLMTagwiseDisp(y, design) fit <- glmFit(y, design) colnames(fit) # DE between untreated and treated for cases lrt <- glmLRT(fit, coef="ConditionCase: Treatments ") # DE between untreated and treated for cases lrt2 <- glmLRT(fit, coef="ConditionControl:Treatments ") # DE between untreated and treated responding differently between cases and controls lrt3 <- glmLRT(fit, contrast=c(0,0,0,0,1,-1) [[alternative HTML version deleted]]