Hi, First, I am very sorry that my attachments are still over size limit. I couldn't find a way to make it smaller and readable at the same time. I got 30 affy arrays (hgu133a) which were carried out over a period of two and half years. There are six types of samples, a,b,c,d,e,f. Each type has different number of biological replicates except type f (no technical replicates). I am going to do pair wise comparisons among these types. For checking array's quality, I read previous posts about this subject and compared my plots with PLM Image Gallery. But how to decide which array is acceptable or not acceptable is still puzzling me. How much difference in these plots across arrays is considered as serious difference? And corresponding arrays should be excluded in down-stream data analysis? sample c2 is obviously different from other arrays in density plot, NUSE plot, and it has large residuals in residual plot and large average background. Maybe it's better to remove this array. But how about the arrays with difference not as big as c2. Please see attachment files. In file "residual.png",I included four arrays'residual images which has bigger artifacts comparing with other arrays. Any comments and suggestions would be greatly appreciated. Thanks! Ren -------------- next part -------------- A non-text attachment was scrubbed... Name: hist_NUSE_RLE.png Type: image/png Size: 7358 bytes Desc: hist_NUSE_RLE.png Url : https://stat.ethz.ch/pipermail/bioconductor/attachments/20060203 /27e1eed9/hist_NUSE_RLE.png -------------- next part -------------- A non-text attachment was scrubbed... Name: residual.png Type: image/png Size: 42943 bytes Desc: residual.png Url : https://stat.ethz.ch/pipermail/bioconductor/attachments/20060203 /27e1eed9/residual.png

Hello Ren, Being puzzled in this case is a good thing as there are no hard and fast rules as to which outlier chips should be excluded from an analysis. In a sense, the answer really depends on the analysis in question. One way to answer the question of which chips should be excluded, or treated differently, is to see the impact that each chip has on the results. You mentioned planning to do pairwise comparisons. Why not initially include all chips and see how aberrant the results are for the pairs that include the outlier chips. If your pairwise comparison entails looking for differentially expressed genes, I suspect that b6 will be the worst offender in spite of having normal looking NUSE values. Hope this helps. -francois --- "Na, Ren" <na at="" uthscsa.edu=""> wrote: > Hi, > > > > First, I am very sorry that my attachments are still > over size limit. > > I couldn't find a way to make it smaller and > readable at the > > same time. > > I got 30 affy arrays (hgu133a) which were carried > out over a period > > of two and half years. There are six types of > samples, a,b,c,d,e,f. > > Each type has different number of biological > replicates except type > > f (no technical replicates). I am going to do pair > wise comparisons > > among these types. > > For checking array's quality, I read previous posts > about this > > subject and compared my plots with PLM Image > Gallery. But how > > to decide which array is acceptable or not > acceptable is still > > puzzling me. How much difference in these plots > across arrays is > > considered as serious difference? And corresponding > arrays should > > be excluded in down-stream data analysis? > > sample c2 is obviously different from other arrays > in density plot, > > NUSE plot, and it has large residuals in residual > plot and large > > average background. Maybe it's better to remove this > array. But how > > about the arrays with difference not as big as c2. > Please see > > attachment files. > > In file "residual.png",I included four > arrays'residual images which > > has bigger artifacts comparing with other arrays. > > > > Any comments and suggestions would be greatly > appreciated. > > > > Thanks! > > > > Ren > > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >

An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20060206/ 1dea0b94/attachment.pl