I am working with 2D gel electrophoresis proteomic datasets and would like to develop or adapt a method for screening out entire gels, after appropriate filtering out of artifacts, normalization, and imputation of missing values, that are outliers. Each gel represents a sample from a different animal and there are times when one animal in a group simply doesn't respond the way others do. This will show up on clustering if there is a clear difference between groups, where the outlier may cluster with the experimental group or, in an extreme case, will form a cluster of 1 all by itself. I would like to approach this numerically and have looked at the Li/Wong dChip approach implanted in BioC as fit.li.wong(affy) but am not sure this model would be applicable for protein gels where there really are no replicates within animal for each protein. On the other hand, perhaps one could consider each set of gel spots as a probeset? Any ideas would be greatly appreciated. Thanks, Mark Mark W. Kimpel MD