Hi, There, I read the vsn vignettes. Although it says that vsn can take a marrayRaw object, there isn't any example that shows how to normalize the marrayRaw object. When I input a marrayRaw object called west2005Raw which is an output of read.SMD, I got the following: > vsn(west2005Raw) Converting marrayRaw (41088 spots, 57 slides) to 41088x114 matrix. Gf-Gb in odd columns, Rf-Rb in even columns. Error in getIntensityMatrix(intensities, verbose) : 'intensities' must not contain NA values. This could indicate that the input data has already undergone some thresholding or transformation (log?), and may not satisfy the requirements of the multiplicative-additive noise model. If you are sure that it is meaningful to proceed, please consider calling vsn on a subset of data where all values are defined, and then use vsnh on the full set of data. It seems that vsn normalizes the Gf-Gb and Rf-Rb seperately, and it doesn't take care of the NAs which directly come from the read.SMD result. Anyone has any experience with this? I am using R 2.2.0 on Linux. Thanks in advavnce! Zhuang Wu

Hi Zhuang, indeed, vsn doesn't like NA values in its input. What you can do here is described in the error message that you quoted, here again a bit more explicit: 1. subselect only those probes that do no have an NA value on any chip 2. apply vsn to that, to estimate the normalization parameters 3. use the function vsnh, with the estimated parameters, to apply the normalization and transformation to the complete set of data. Best wishes Wolfgang Wu, Zhuang wrote: > Hi, There, > > I read the vsn vignettes. Although it says that vsn can take a marrayRaw object, > there isn't any example that shows how to normalize the marrayRaw object. When I > input a marrayRaw object called west2005Raw which is an output of read.SMD, I > got the following: > >> vsn(west2005Raw) > Converting marrayRaw (41088 spots, 57 slides) to 41088x114 matrix. > Gf-Gb in odd columns, Rf-Rb in even columns. > Error in getIntensityMatrix(intensities, verbose) : > 'intensities' must not contain NA values. > This could indicate that the input data has already undergone some > thresholding or transformation (log?), and may not satisfy the > requirements of the multiplicative-additive noise model. > If you are sure that it is meaningful to proceed, please > consider calling vsn on a subset of data where all values > are defined, and then use vsnh on the full set of data. > > It seems that vsn normalizes the Gf-Gb and Rf-Rb seperately, and it doesn't take > care of the NAs which directly come from the read.SMD result. Anyone has any > experience with this? > > I am using R 2.2.0 on Linux. > > Thanks in advavnce! > > Zhuang Wu > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor