We should keep in mind that systematic differences detected among probe intensities within a probe sets are tiny in comparison to the typical variation in intensity observed among probes within a probe set. Systematic differences are detectable because we are summarizing probe intensities by location (3' -> 5' order) over 20 000+ probes (in HG U133 case). A trend from 3' to 5' order in average probe intensity is expected (look at the Affymetrix HG 133A spike in experiment for an example of how the digestion plots would look like in the ideal case) . This trend varies from chip type to chip type. It is not clear to me how deviations from the expected trend translates into impact on expression summaries obtained for the particular chips where these deviations are observed. One observation that I have made is that poor quality chips (with "quality" ascertained by examination of chip image, or summaries of RMA residuals, or variability in log ratios of expressions) are very often characterized with an absence of 3'-5' trend in RNA digestion plot. If you look at the RNA digestion plots to judge chip quality, beware of deviations from expected 3'-5' trend in both directions (absence of trend being more telling of potential problems in my experience). -francois Lawrence Paul Petalidis <lpp22@cam.ac.uk> wrote: Hi William The RNA plots on your site look great to me. I routinely work with tumour clinical RNA and the slopes are significantly larger than yours even if the quality of starting RNA appears to be good. You see, RNA quality control relies heavily on ribosomal RNA quality (peaks). However, I have had samples with great ribosomal peaks that actually produce RNAdeg plots with a significant slope 5' to 3'. Its the state of the mRNA that counts and that seems to be more sensitive to degradation than the ribosomal one. I would be very happy working with the ones you show on your site. With kind regards, Lawrence ______________________________ Lawrence Paul Petalidis Ph.D. Candidate University of Cambridge Department of Pathology ______________________________ -----Original Message----- From: bioconductor-bounces@stat.math.ethz.ch [mailto:bioconductor-bounces@stat.math.ethz.ch]On Behalf Of William Kenworthy Sent: 20 November 2003 07:22 To: BioConductor List Subject: Re: [BioC] RNA digestion plots RNA plots are something I have also been looking at, but all the datasets I have access to can be characterised as "good". Can anyone supply a (anonymous?) "cel" file of any persuasion that is known to have been used with degraded RNA? On larger arrays, I do see a slight slope, but not any "rolloff" which I think would represent degradation. At the present time, I add a "typical" CELfile from another project that seems good and adding that to the current affybatch and running the plots. Currently, they track very well across projects (especially if normalised) - so a sample "bad one" would be nice just to see what it looks like! I have put four plots up at "http://wdk.dyndns.org/rna/rnadeg.html". The red and black traces represent samples that I was informed "might" have been questionable, but they did pass affy's check criteria, the green trace is from an experiment that was considered good. BillK On Thu, 2003-11-20 at 13:27, Kevin Dawson wrote: > Does any one know when to take the RNA digestion plots seriously. On > hgu133a arrays, these plots always seem to incline from 5' to 3'. Is there > a certain slope value that sets the good and bad apart? > > Thank you > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor [[alternative HTML version deleted]]