dear ALL: i have 35 samples to find DE genes between "c0h" vs t0h", i can use such coding to reach the same aim, what is the difference? which one is better? raw.data <- read.table("expression- table.txt",row.names=1)#?1??????,??????????? lib_size <- read.table("lib_size.txt"); lib_size <- unlist(lib_size) #??????????? d <- raw.data[, 2:dim(raw.data)[2]]#?1?????????????????? length<-raw.data[, 1] method 1: group =factor(c('c0h','c0h','c0h','c24h','c24h','c24h','t0h','t0h','t0 h','t6h','t6h','t6h','t6h','t12h','t12h','t12h','t12h','t18h','t18h',' t18h','t18h', 't24h','t24h','t24h','t36h','t36h','t36h','t48h','t48h',' t48h','c6h','c12h','c18h','c36h','c48h')) d <- DGEList(counts = d, lib.size = lib_size, group = group) d <- calcNormFactors(d) #To estimate common dispersion: d <- estimateCommonDisp(d) #To estimate tagwise dispersions: d <- estimateTagwiseDisp(d) et <- exactTest(d, pair=c("c0h","t0h"), dispersion="tagwise") sig1<-summary(de <- decideTestsDGE(et, p=0.05, adjust="BH")) sig1<-as.vector(sig1) result <- topTags(et, n=dim(d)[1], adjust.method="BH", sort.by="p.value") write.table(result,file = "t0h_c0h",sep = "\t") method 2: #??DGEList?? d <- DGEList(counts = d, lib.size = lib_size) #normalization dge <- calcNormFactors(d) rm(d) fac=factor(c('c0h','c0h','c0h','c24h','c24h','c24h','t0h','t0h','t0h', 't6h','t6h','t6h','t6h','t12h','t12h','t12h','t12h','t18h','t18h','t18 h','t18h', 't24h','t24h','t24h','t36h','t36h','t36h','t48h','t48h',' t48h','c6h','c12h','c18h','c36h','c48h')) design = model.matrix(~fac) colnames(design) <- levels(fac) dge <- estimateGLMCommonDisp(dge, design) dge <- estimateGLMTagwiseDisp(dge, design) glmfit.dge <- glmFit(dge, design, dispersion=dge$tagwise.dispersion) t0_c0 = makeContrasts("t0h-c0h",levels=design) lrt.dge <- glmLRT(dge, glmfit.dge, contrast = t0_c0) result <- topTags(lrt.dge, n=dim(dge)[1], adjust.method="BH", sort.by="p.value") write.table(result,file = "t0_c0",sep = "\t") -- shan gao Room 231(Dr.Fei lab) Boyce Thompson Institute Cornell University Tower Road, Ithaca, NY 14853-1801 Office phone: 1-607-254-1267(day) Official email:sg839 at cornell.edu Facebook:http://www.facebook.com/profile.php?id=100001986532253

>dear ALL: > >i have 35 samples to find DE genes between "c0h" vs t0h", i can use >such coding to reach the same >aim, what is the difference? which one is better? The linear model is better than the exact test because you have a time course, which means the same biological organism measured repeatedly. You can't model organism-specific effects using the exact test.