Dear list members, in chapter 4 of "Textual description of affy" several methods for quality control are described: hist(), image(), boxplot() and RNA degradation plots. Has anybody used these methods to identify bad quality chips? When do you exclude a chip from further analysis? What's your decision rule? Any other suggestions on how to identify bad quality from the data alone? Thanks a lot. -oli- -- Oliver Hartmann, Institute of Medical Biometry and Epidemiology Philipps-University Marburg, Bunsenstr. 3, D-35037 Marburg phone +49(0)6421 28 66514, fax +49(0)6421 28 68921

A lot of target sequences are very short and just in 3'UTR, does RNA degradation plot really show overall degradation? Xueqing -----Original Message----- From: Oliver Hartmann [mailto:hartmann@mailer.uni-marburg.de] Sent: Thursday, February 13, 2003 11:59 AM To: bioconductor@stat.math.ethz.ch Subject: [BioC] Quality of Affy Chips with the affy package Dear list members, in chapter 4 of "Textual description of affy" several methods for quality control are described: hist(), image(), boxplot() and RNA degradation plots. Has anybody used these methods to identify bad quality chips? When do you exclude a chip from further analysis? What's your decision rule? Any other suggestions on how to identify bad quality from the data alone? Thanks a lot. -oli- -- Oliver Hartmann, Institute of Medical Biometry and Epidemiology Philipps-University Marburg, Bunsenstr. 3, D-35037 Marburg phone +49(0)6421 28 66514, fax +49(0)6421 28 68921 _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor

The probe sets have been designed to be within the workable 3' amplified region under normal circumstances. It maybe worth examining particularly when you are working with small RNA samples and doing 2 rounds of amplification. -----Original Message----- From: Zhang, Xueqing To: bioconductor Sent: 13/02/03 17:24 Subject: RE: [BioC] Quality of Affy Chips with the affy package A lot of target sequences are very short and just in 3'UTR, does RNA degradation plot really show overall degradation? Xueqing -----Original Message----- From: Oliver Hartmann [mailto:hartmann@mailer.uni-marburg.de] Sent: Thursday, February 13, 2003 11:59 AM To: bioconductor@stat.math.ethz.ch Subject: [BioC] Quality of Affy Chips with the affy package Dear list members, in chapter 4 of "Textual description of affy" several methods for quality control are described: hist(), image(), boxplot() and RNA degradation plots. Has anybody used these methods to identify bad quality chips? When do you exclude a chip from further analysis? What's your decision rule? Any other suggestions on how to identify bad quality from the data alone? Thanks a lot. -oli- -- Oliver Hartmann, Institute of Medical Biometry and Epidemiology Philipps-University Marburg, Bunsenstr. 3, D-35037 Marburg phone +49(0)6421 28 66514, fax +49(0)6421 28 68921 _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor ********************************************************************** This email and any files transmitted with it are confidential an ... [[dropped]]

On Thu, Feb 13, 2003 at 05:58:39PM +0100, Oliver Hartmann wrote: > Dear list members, > > in chapter 4 of "Textual description of affy" several methods for > quality control are described: hist(), image(), boxplot() and RNA > degradation plots. > > Has anybody used these methods to identify bad quality chips? When do > you exclude a chip from further analysis? What's your decision rule? > Any other suggestions on how to identify bad quality from the data alone? 'image' lets you do a visual inspection of the chip (scratches, air bubble(s), washing step problem, ..., should be visible). 'hist' can let you see saturation (if you have the distribution of probe intensities showing a strange 'bump' on the higher end of the intensity range, you may suspect that the captor (a CCD camera I believe) get too much an cut off the signal. You can confirm with a scatter plot (or matrix of scatter plots (see pairs.AffyBatch))). 'boxplot' and 'hist' can let you vizualize the effect of normalization for example (quality control on the processing step). Few other things are being worked on... About the decision rule, I thought it was a function of parameters 'cost', 'deadline', 'availability of samples' and 'results'. ;) L.

it's not clear there is a protocol for excluding arrays. as laurent pointed out, different labs/companies may have different guidelines. i can give you two anecdotes of my own: 1- in one specific batch i analyzed i excluded one array (out of 59). a scatter plot of log intensites with its replicates showed shapes very different from a straight line with scatter. none of the normalization procedures fixed it. 2-i had a set of arrays where for some genes i was getting extremely small within group variances. aftere looking at the histograms i noticed there was lots of staturation. the genes having this characterisitc happened to be higly expressed genes. in this particular case i was very careful when looking at across-condition-to-within-condition ratios of variances type tests (e.g. t-test, F-test). by talking to people who analyze lots of arrays im sure you can learn from their anecdotes. hope this helps, rafael On Thu, 13 Feb 2003, Oliver Hartmann wrote: > Dear list members, > > in chapter 4 of "Textual description of affy" several methods for > quality control are described: hist(), image(), boxplot() and RNA > degradation plots. > > Has anybody used these methods to identify bad quality chips? When do > you exclude a chip from further analysis? What's your decision rule? > Any other suggestions on how to identify bad quality from the data alone? > > Thanks a lot. > > -oli- > -- > Oliver Hartmann, Institute of Medical Biometry and Epidemiology > Philipps-University Marburg, Bunsenstr. 3, D-35037 Marburg > phone +49(0)6421 28 66514, fax +49(0)6421 28 68921 > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >

Dchip has a facility to detect outliers of a group of arrays--i.e. where the probe pm-mm patterns don't correlate with the group. This is often due to scratches etc.. See http://biosun1.harvard.edu/complab/dchip/data%20view.htm#outlier These probesets can be exported and I imagine then used to exclude comparisons within bioconductor. See http://biosun1.harvard.edu/complab/dchip/model- based%20expression.htm#export _outlier -----Original Message----- From: Hinnerk Boriss [mailto:boriss@izbi.uni-leipzig.de] Sent: Thursday, February 20, 2003 5:26 PM To: 'Laurent Gautier' Cc: bioconductor Subject: RE: [BioC] Quality of Affy Chips with the affy package > 'image' lets you do a visual inspection of the chip > (scratches, air bubble(s), > washing step problem, ..., should be visible). > 'hist' can let you see saturation (if you have the > distribution of probe > intensities showing a strange 'bump' on the higher end of the > intensity range, > you may suspect that the captor (a CCD camera I believe) get too much > an cut off the signal. You can confirm with a scatter plot (or matrix > of scatter plots (see pairs.AffyBatch))). > 'boxplot' and 'hist' can let you vizualize the effect of normalization > for example (quality control on the processing step). > Hi, if I find such areas with scratches, bubbles or saturation on a chip, how can I mask them? Also, if probe level intensities go into saturation, how does that affect signal values? I ask, because there will generally be probes in a probe set that won't have that problem. Sure, this will depend on the signal value estimation method. How do RMA and MBEI react when some probes ru in saturation? Best, Hinnerk _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor ********************************************************************** This email and any files transmitted with it are confidential an ... [[dropped]]