Dear all, Recently we have performed a microarray time-scale experiment including different time-points (0h, 2h, 6h, ). Treated and untreated cells were compared by competitive hybridizations for every time-point using dual-channel microarrays. Every time-point was performed in triplicate including 3 biological replicates. Microarray analysis is performed in R with Limma using linear models (but not based on a separate channel analysis (mixed models)!!!). Comparing the boxplots for the three biological replicates on every time-point it seems as the data distribution is quite different between the replicates. In such a case, if I understood right it might sometimes be worth to perform between array normalization by performing scale normalization between the three replicates for every given time-point. How could this be performed best? By splitting up the MAList object into as many MAList objects as there are different time-points (always grouping together the three replicates for a time-point into one object) before performing between array normalization and then merging it back afterwards into a single object? Are there other possibilities to perform this? Is it correct to perform between array normalization for the three replicates on every time-point as I described here? Here you may find the corresponding R code: >#Splitting up MAList object (MA): >MA.t0 = MA[, 1:3] >MA.t2 = MA[, 4:6] >MA.t6 = MA[, 7:9] > >#Between array scale normalization for the different time-points?: >MA.t0 = normalizeBetweenArrays(MA.t0, method="scale") >MA.t2 = normalizeBetweenArrays(MA.t2, method="scale") >MA.t6 = normalizeBetweenArrays(MA.t6, method="scale") > >#Merging normalized data back into a single object for further analysis: >MA = cbind(MA.t0, MA.t2, MA.t6) I would appreciate any comments or suggestions. Regards, Serge Eifes Serge Eifes Laboratoire de Biologie Moleculaire et Cellulaire du Cancer (LBMCC) Hopital Kirchberg 9,rue Edward steichen L-2540 LUXEMBOURG Phone:+ 352 2468-4046 Fax : + 352 2468-4060

Dear Serge, > Recently we have performed a microarray time-scale experiment including > different time-points (0h, 2h, 6h, ...). Treated and untreated cells were > compared by competitive hybridizations for every time-point using > dual-channel microarrays. Every time-point was performed in triplicate > including 3 biological replicates. > > Comparing the boxplots for the three biological replicates on every > time-point it seems as the data distribution is quite different between the > replicates. In such a case, if I understood right it might sometimes be > worth to perform between array normalization by .... Since you say "replicates"...: Another action that might be worthwhile in such a case is to carefully look at scatterplots, histogrammes, spatial plots of the data and to carefully consider the question whether the data quality is appropriate for the intended analysis. Best wishes Wolfgang ------------------------------------------------------------------ Wolfgang Huber EBI/EMBL Cambridge UK http://www.ebi.ac.uk/huber