Hi, I have analyzed a data set with 2 different classes UM and M(with subcklasses M1 and M2) . I have fitted the linear model with limma for the coefficients UM, M1 and M2 and I have compared UM vs (M1+M2).I found a significant change (adjuste p-value<0.0001 and B>2) for 236 genes I did the analysis also with SAM (with the function samrocNboot in the package SAGx)comparing UM vs M.I found a significant change(adjusted p-value <0.001) for 285 genes. I had also 29 genes in common between the two anlalysis. For visualization pouposes for both results I used, on normalized data matrix, a hierarchical clustering (with pearson correlation as distance and average as method). But with the SAM's genes I obtained a good clustering, with a good separation between the two classes. For LIMMA's genes I couldn't succed to obtain a good separation between the two classes. Have you any idea about? May be is SAM closer to a mesure of correlation, withou fitting any linear model, than LIMMA? Thanks for any suggestion Daniela

daniela marconi wrote: > Hi, I have analyzed a data set with 2 different classes UM and M(with > subcklasses M1 and M2) . I have fitted the linear model with limma > for the coefficients UM, M1 and M2 and I have compared UM vs > (M1+M2).I found a significant change (adjuste p-value<0.0001 and B>2) When you say you compared UM vs (M1+M2), is that what you used in your call to makeContrasts() (e.g., makeContrasts(UM - (M1 + M2))? If so, you are comparing UM to the *sum* of M1 and M2 instead of the *mean* of M1 and M2, which would probably explain the differences. Best, Jim > for 236 genes I did the analysis also with SAM (with the function > samrocNboot in the package SAGx)comparing UM vs M.I found a > significant change(adjusted p-value <0.001) for 285 genes. > > I had also 29 genes in common between the two anlalysis. > > For visualization pouposes for both results I used, on normalized > data matrix, a hierarchical clustering (with pearson correlation as > distance and average as method). But with the SAM's genes I obtained > a good clustering, with a good separation between the two classes. > For LIMMA's genes I couldn't succed to obtain a good separation > between the two classes. Have you any idea about? May be is SAM > closer to a mesure of correlation, withou fitting any linear model, > than LIMMA? Thanks for any suggestion Daniela > > _______________________________________________ Bioconductor mailing > list Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor -- James W. MacDonald Affymetrix and cDNA Microarray Core University of Michigan Cancer Center 1500 E. Medical Center Drive 7410 CCGC Ann Arbor MI 48109 734-647-5623