Dear Alessio, if "isBlank" is a logical vector indicating which spots are blank, and x is an RGList, exprSet or a matrix, you can simply use xr <- x[!isBlank, ] and work on xr. If the number of blank spots is not large, they should not make a big difference to vsn, since it uses a robust algorithm that tries to be insensitive to outliers (of course removing known outliers is always better than hoping for an algorithm to do it.) The "subsample" argument is for random subsampling, and is not intended for what you want to do. HTH Wolfgang Alessio Venier wrote: > Hi, > > I'm using the limma package (2.7.5) to analyze data from spotted microarray, I've used the read.maimage function > to load the raw data from files generated by bluefuse (this files also contain intensities from blank spots). > Vesion 2.7.5 is working fine now when loading more than one file per time (thanks to Gordon K Smyth and Mike Smith). > > I would like to use the vsn normalization method (I've installed the vsn R package). > > I'm wondering if the intensity values from blank spots could create problems with normalization process from > a mathematical point of view when applying vsn normalization method. > > If it's the case I would like to "filter" the blank spots using R instead of bluefuse features. > > I've tried to assign NA values to all the blank spots intensities for all channels/arrays but when I try to use > the vsn function I receive a message error becouse this function doesn't allow NA values in in the data > to be analyzed. I don't understand how to use the subsample argument in vsn to solve my problem. > It requires an integer value to define a subsample but I would require a logic value > (something like: use only non NA values). > > Any advice? > > Kind Regards > Alessio >