Dear all, I have begun working with some affy data and there are some aspects of it that are confusing to me. The data was analysed using MAS 5.0 suite and most of the questions here pertain to single arrays (aka Absolute Analysis?). The first 3 question is on Scaling Factors (SF) which is defined as SF = TGT / x_96 ; where x_96 is the 2% trimmed mean 1. Why does MAS 5.0 apply the calculated SF to the middle 96% only and not 100% of the data ? Since it is often that SF > 1 and signal > 0 by affy's definition, the top 2% are being penalized. What is the rational behind this ? 2. I have read that a SF > 3 generally implies problems with the experiment. Why is this? 3. How is the target intensity, TGT, chosen ? Values of 100, 200, 500 and even 1500 seem to have been chosen. Or do we choose TGT to ensure SF < 3 for most of the arrays. The next 2 are related to the Present/Absent calls provided. 4. I have 74 arrays and on average 50% of the data is missing. What is the normal/acceptable observed percentage missing. 5. I am removing all the signal data with Absent Call. Is this correct and is there any other use for Absent Call. Googling and reading the affy manuals (some parts of it anyway) has not cleared these confusions. Any answers, advices or references from this list is much appreciated. Thank you for your time. Regards, Adai.

On Fri, Apr 18, 2003 at 11:36:39AM +0800, Adaikalavan Ramasamy wrote: > Dear all, > > I have begun working with some affy data and there are some aspects of > it that are confusing to me. The data was analysed using MAS 5.0 suite > and most of the questions here pertain to single arrays (aka Absolute > Analysis?). Do you mean that you use the MAS software ? If this is the case, we might be in the same position as you (i.e. with the official technical documentation provided, especially the sadd_white-paper.pdf that can be found on the website of the company). Did you read this document ? Regards, L. > > The first 3 question is on Scaling Factors (SF) which is defined as > > SF = TGT / x_96 ; where x_96 is the 2% trimmed mean > > > 1. Why does MAS 5.0 apply the calculated SF to the middle 96% only and > not 100% of the data ? Since it is often that SF > 1 and signal > 0 by > affy's definition, the top 2% are being penalized. What is the rational > behind this ? > > 2. I have read that a SF > 3 generally implies problems with the > experiment. Why is this? > > 3. How is the target intensity, TGT, chosen ? Values of 100, 200, 500 > and even 1500 seem to have been chosen. Or do we choose TGT to ensure SF > < 3 for most of the arrays. > > > The next 2 are related to the Present/Absent calls provided. > > 4. I have 74 arrays and on average 50% of the data is missing. What is > the normal/acceptable observed percentage missing. > > 5. I am removing all the signal data with Absent Call. Is this correct > and is there any other use for Absent Call. > > > Googling and reading the affy manuals (some parts of it anyway) has not > cleared these confusions. Any answers, advices or references from this > list is much appreciated. Thank you for your time. > > Regards, Adai. > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor

Dear Laurent, Thank you for the reference. I have just finishing reading it and I think I am more confused because on page 8, it appears that they intended to multiply the scaling factor to 100% of the data. " The reported value of probe set i is: ReportedValue(i) = nf * sf * 2 ^ SignalLogValue " My data is contrary to this. If possible, could you check the mean value and the 2% trimmed mean after applying to scale factor to eliminate the possibility error on my part. Much appreciated. Can anyone shed some light on the remaining questions ? Or is there an affy mailing list group I can ask. Thanks. Regards, Adai. -----Original Message----- From: Laurent Gautier [mailto:laurent@cbs.dtu.dk] Sent: Friday, April 18, 2003 11:57 AM To: Adaikalavan Ramasamy Cc: Bioconductor mailing list Subject: Re: [BioC] OFF-TOPIC: Affy Scaling Factor and Absent/Present Calls On Fri, Apr 18, 2003 at 11:36:39AM +0800, Adaikalavan Ramasamy wrote: > Dear all, > > I have begun working with some affy data and there are some aspects of > it that are confusing to me. The data was analysed using MAS 5.0 suite > and most of the questions here pertain to single arrays (aka Absolute > Analysis?). Do you mean that you use the MAS software ? If this is the case, we might be in the same position as you (i.e. with the official technical documentation provided, especially the sadd_white-paper.pdf that can be found on the website of the company). Did you read this document ? Regards, L. > > The first 3 question is on Scaling Factors (SF) which is defined as > > SF = TGT / x_96 ; where x_96 is the 2% trimmed mean > > > 1. Why does MAS 5.0 apply the calculated SF to the middle 96% only and > not 100% of the data ? Since it is often that SF > 1 and signal > 0 by > affy's definition, the top 2% are being penalized. What is the > rational behind this ? > > 2. I have read that a SF > 3 generally implies problems with the > experiment. Why is this? > > 3. How is the target intensity, TGT, chosen ? Values of 100, 200, 500 > and even 1500 seem to have been chosen. Or do we choose TGT to ensure > SF < 3 for most of the arrays. > > > The next 2 are related to the Present/Absent calls provided. > > 4. I have 74 arrays and on average 50% of the data is missing. What is > the normal/acceptable observed percentage missing. > > 5. I am removing all the signal data with Absent Call. Is this correct > and is there any other use for Absent Call. > > > Googling and reading the affy manuals (some parts of it anyway) has > not cleared these confusions. Any answers, advices or references from > this list is much appreciated. Thank you for your time. > > Regards, Adai. > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor