In the "oligo" manual there is a section that describes how to calculate all genes with 2x change or more. This sections uses rowMeans to calculate the average of replicates-"rowMeans(e[, index])". However, since the expression values in eset are in log2, is rowMeans the correct way to calculate averages? I thought rowmeans was for calculating the averages of normal numbers, not log2 numbers. If this method is not correct, how should the averages of the replicates be calculated? Thanks, Mary A. Allen from the manual 2.2.4 Assessing dif erential expression One simple approach to assess dif erential expression is to flag units with log-ratios greater (in absolute value) than 1, i.e. a change greater than 2-fold when comparing brain vs. universal reference. R> e <- exprs(eset) R> index <- which(eset[["Key"]] == "brain") R> d <- rowMeans(e[, index])-rowMeans(e[, -index]) R> a <- rowMeans(e) R> sum(abs(d)>1) [1] 10043

Just to make it clear for others: what Mary describes is just one portion of the section... and I know (and I show later on that document) that there are other (preferable) ways to assess differential expression... Mary, there isn't such distinction: rowMeans() is simply estimating for you the average log-expression for each of the conditions (so, you get the average log-expression for gene1, gene2, ..., geneN). To get averages, the normality assumption doesn't play a role. b On 15 November 2012 22:12, Mary Ann Allen <mary.a.allen@colorado.edu> wrote: > In the "oligo" manual there is a section that describes how to calculate > all genes with 2x change or more. This sections uses rowMeans to calculate > the average of replicates-"rowMeans(e[, index])". However, since the > expression values in eset are in log2, is rowMeans the correct way to > calculate averages? I thought rowmeans was for calculating the averages of > normal numbers, not log2 numbers. If this method is not correct, how should > the averages of the replicates be calculated? > > Thanks, > Mary A. Allen > > > from the manual > > 2.2.4 Assessing dif erential expression > One simple approach to assess dif erential expression is to flag units > with log-ratios greater (in absolute value) than 1, i.e. a change greater > than 2-fold when comparing brain vs. universal reference. > R> e <- exprs(eset) > R> index <- which(eset[["Key"]] == "brain") > R> d <- rowMeans(e[, index])-rowMeans(e[, -index]) > R> a <- rowMeans(e) > R> sum(abs(d)>1) > [1] 10043 > > ______________________________**_________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/**listinfo/bioconductor<https: stat.et="" hz.ch="" mailman="" listinfo="" bioconductor=""> > Search the archives: http://news.gmane.org/gmane.** > science.biology.informatics.**conductor<http: news.gmane.org="" gmane.="" science.biology.informatics.conductor=""> > [[alternative HTML version deleted]]