Howdy, Just some thoughts: On Wed, Apr 6, 2011 at 11:59 AM, James W. MacDonald <jmacdon at="" med.umich.edu=""> wrote: > Hi Seraya, > > On 4/6/2011 11:16 AM, Seraya Maouche wrote: [snip] > I think the reviewer's point is that you are computing a fold-change based > on something that is primarily due to signal divided by something that is > primarily due to noise. In other words, the numerator is signal, and the > denominator is noise. > > In addition, the value due primarily to noise is very close to zero, so you > end up with a huge fold change that can vary widely, depending on the > variability of your noise signal. So the actual value of the ratio is > probably not that meaningful, but the fact that it is reliably large > probably means that the gene is truly differentially expressed. You just > cannot say reliably by how much. I don't know ... It's not entirely clear to me what is causing the reviewer to be so upset about this. Seraya, is there some assertion you are making regarding the actual *value* of the fold change in these cases (where the denominator is "absent"), or are you just saying that these (this) gene is differentially express in cond_a vs. cond_b? Anyway, in order to sidestep this "technicality", maybe one could put the fold change into an "appropriate" scale after-the-fact? For instance, in these situations where the gene is absent in one expt, and not in the other, maybe one could set the expression of the "absent" probes/genes in a given condition to be the minimum value of a detected/expressed gene in that condition. Then calculate the fold change post-facto? Where by post-facto I mean to just let limma do "it's thing" without any data manipulation. Then, when Seraya is listing some fold change value (in a table in the paper(?)), replace the fold change of XX/0 (= Inf) with XX/(min gene expression in expt the gene is absent in)? I mean, the whole thing is a bit strange to me, unless you are indeed asserting something based on the value of the fold change -- in which case, the value itself in these situations I guess doesn't make much sense at all and I reckon I can see the point the reviewer is nitpicking over (?). Just a thought, -steve -- Steve Lianoglou Graduate Student: Computational Systems Biology ?| Memorial Sloan-Kettering Cancer Center ?| Weill Medical College of Cornell University Contact Info: http://cbio.mskcc.org/~lianos/contact

Hi all, I think it's helpful to distinguish, conceptually, between the true state of nature (gene not expressed, infinite fold change from 0 to some finite concentration) and our imperfect measurement (gene not detected, very large observed fluorescence intensity ratio resulting from dividing a large 'signal' value by a small background 'noise' value). This is really nothing specific to microarrays or two colours, it applies to any physical measurement. The reviewer seems to be upset since the ms conflates these concepts and thus makes confusing statements. Best wishes Wolfgang Il Apr/7/11 1:02 AM, Wei Shi ha scritto: > Hi Seraya: > > It is true for microarray as well that the signal could have a quantity even it corresponds to noise not to the real signal. This is because during normalization an offset is often added to the data (before log transformation) either explicitly or implicitly to improve the precision (reduce variances between replicates). This is also the reason why you would never see probes with zero intensity (raw scale) in a microarray analysis. Maybe you should make it clear to the reviewer that absent/non-expressed probes have intensities too. This is the way how microarray data are being analyzed. > > Illumina calculate the detection p values of probes using the intensities of all the negative control probes available on the array (~1000 such probes). For example, if a probe has a detection p value of 0.05, this will mean only 5% of all the negative control probes have intensities larger than the intensity of that probe. Maybe this will help you to explain how you determined whether a gene was expressed or not. > > Cheers, > Wei > > On Apr 7, 2011, at 1:16 AM, Seraya Maouche wrote: > >> Dear Jenny, >> Thank you, >> I agree that "absence" does not mean that the transcript is not expressed >> because if the probe sequence used to target that transcript is not >> performing well so non expression is the result of the technical problem >> rather than the abundance of the transcript in the sample analyzed. We can >> also have other factors such image acquisition conditions, etc. >> >> But for my analyses, as I have a large sample size (n> 1500), if a >> transcript is absent in all RNA samples, we cannot expect that this is just >> an artifact. >> What I am doing is calculating detection calls and then I have three >> situation: 1) probes present in all samples, 2) probes absent in all >> samples, and 3) probes absent in a subset of samples. In order to not lose >> information, I am not filtering probes of category (3) >> >> And I also examine the concordance of detection call for all probes >> (spliced variants or technical replicate probes) tagging the same gene, so >> if I have 3 probes >> for the same gene, all absent I consider this as confidence to say that the >> gene is not expressed. >> >> In addition, I am using other gene expression datasets (for the same cell >> type), generated using different platforms to check whether the probe absent >> in dataset A is also absent in dataset B, C.. >> >> Now for the reviewer, he cannot understand how we say that a gene is not >> expressed and on the other hand, we use the intensity for this gene to >> calculate a fold change. But in signal processing, we can have a quantity of >> signal even it corresponds to noise and not to the real signal. >> >> Best wishes, >> Seraya >> >> >> -----Original Message----- >> From: Jenny Drnevich [mailto:drnevich at illinois.edu] >> Sent: Mittwoch, 6. April 2011 16:53 >> To: Seraya Maouche; 'James W. MacDonald'; 'Wei Shi' >> Cc: Bioconductor at r-project.org >> Subject: Re: [BioC] Limma >> >> Hi Seraya, >> >> I think your explanation for the reviewer is on track. Despite the fact that >> some microarray platforms allow the calculation of a metric called "present" >> or "absent", they really cannot detect whether a gene is truly expressed or >> not. >> Instead, it's just a metric saying if the signal was above some noise >> threshold. Now, the signal measured will always be non-zero, so we can use >> these numbers to calculate fold-change regardless of the "present" or >> "absent" call. You are right to throw out only those genes that are called >> "absent" in all samples, but for the rest of the genes, the "present/absent" >> metric is not good enough to categorize "off" versus "on" genes, so we just >> use the numbers measured, calculate fold-changes and do statistical tests. >> Hopefully the reviewer will be satisfied with an explanation of this sort. >> >> Good luck, >> Jenny >> >> At 09:39 AM 4/6/2011, Seraya Maouche wrote: >>> Dear Jim, dear Wei, >>> Thanks for your help, it is not a two color analysis, it is Illumina. >>> >>> Best wishes >>> seraya >>> >>> -----Original Message----- >>> From: James W. MacDonald [mailto:jmacdon at med.umich.edu] >>> Sent: Mittwoch, 6. April 2011 15:55 >>> To: Wei Shi >>> Cc: Seraya Maouche; Bioconductor at r-project.org >>> Subject: Re: [BioC] Limma >>> >>> Hi Wei, >>> >>> I think you misunderstood the OP. Seraya _didn't_ remove genes that >>> were only present in one condition. The problem is that the reviewer >>> didn't like ratios with a zero in the denominator, which is a fair >> complaint. >>> >>> I don't do two color analyses, so don't know what the consensus is for >>> handling logratios where the denominator is really close to zero. Since >>> you guys do this stuff all the time, perhaps you have some pointers? >>> >>> Best, >>> >>> Jim >>> >>> >>> >>> On 4/5/2011 6:43 PM, Wei Shi wrote: >>>> Hi Seraya: >>>> >>>> Genes which are present in one condition but not in the other >>>> should >>> NOT be removed from your analysis. Only those gene which are absent in >>> both conditions should be filtered out to improve the power to detect >>> differentially expressed genes. It is very likely that a lot of genes >>> of biological interest were not included in your analysis results due >>> to the removal of genes which are present in one condition but not in the >> other. >>> Have a look at the case study for processing Illumina BeadChip data in >>> limma user guide about the probe filtering. >>>> >>>> Hope this helps. >>>> >>>> Cheers, >>>> Wei >>>> >>>> On Apr 6, 2011, at 2:05 AM, Seraya Maouche wrote: >>>> >>>>> Dear Prof Gordon, dear Bioconductor members, >>>>> >>>>> I have performed gene expression analysis using Limma (Illumina >>>>> human >>>>> ref8) comparing two types of cells (referred below as cond1 and cond2). >>>>> Based on detection call, I filtered out transcripts which are >>>>> absent in both types of cells. Transcripts which were expressed >>>>> only in one cell type were included in the analysis. >>>>> >>>>> I have received the comment below from a reviewer who seems not >>>>> agree to calculate fold change for genes expressed only in one >> condition. >>>>> Would it be possible to have your opinion about this. >>>>> >>>>> Thank you in advance for your time, S Maouche >>>>> >>>>> "There is a little conceptual difficulty related to the cond1/cond2 >>>>> comparisons for genes that are considered not detected. If a gene >>>>> product is absent (0) in one cell then no fold change can be >>>>> computed (table 2). I don?t know how to circumvent this difficult >>>>> except by saying that the ?noise? is considered to reflect low >>>>> expression. The terms ?not detected? and ?not expressed? are often >>>>> used interchangeably while this is not the same. Detection is based >>>>> on the definition adopted and in many places of the manuscript it >>>>> should be used in place of expression." >>>>> >>>>> >>>>> >>>>> Universit?tsklinikum Schleswig-Holstein Rechtsf?hige Anstalt des >>>>> ?ffentlichen Rechts der Christian-Albrechts-Universit?t zu Kiel und >>>>> der Universit?t zu L?beck >>>>> >>>>> Vorstandsmitglieder: Prof. Dr. Jens Scholz (Vorsitzender), Peter >>>>> Pansegrau, Christa Meyer Vorsitzende des Aufsichtsrates: Dr. >>>>> Cordelia Andre?en >>>>> Bankverbindungen: F?rde Sparkasse BLZ 210 501 70 Kto.-Nr. 100 206, >>>>> Commerzbank AG BLZ 230 800 40 Kto.-Nr. 300 041 200 >>>>> >>>>> Diese E-Mail enth?lt vertrauliche Informationen und ist nur f?r die >>>>> Personen bestimmt, an welche sie gerichtet ist. Sollten Sie nicht >>>>> der bestimmungsgem??e Empf?nger sein, bitten wir Sie, uns hiervon >>>>> unverz?glich zu unterrichten und die E-Mail zu vernichten. >>>>> Wir weisen darauf hin, dass der Gebrauch und die Weiterleitung >>>>> einer nicht bestimmungsgem?? empfangenen E-Mail und ihres Inhalts >>>>> gesetzlich verboten sind und ggf. Schadensersatzanspr?che ausl?sen >>> k?nnen. >>>>> _______________________________________________ >>>>> Bioconductor mailing list >>>>> Bioconductor at r-project.org >>>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>>>> Search the archives: >>>>> http://news.gmane.org/gmane.science.biology.informatics.conductor >>>> >>>> >>>> ____________________________________________________________________ >>>> __ The information in this email is confidential and >>>> intend...{{dropped:6}} >>>> >>>> _______________________________________________ >>>> Bioconductor mailing list >>>> Bioconductor at r-project.org >>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>>> Search the archives: >>>> http://news.gmane.org/gmane.science.biology.informatics.conductor >>> >>> -- >>> James W. MacDonald, M.S. >>> Biostatistician >>> Douglas Lab >>> University of Michigan >>> Department of Human Genetics >>> 5912 Buhl >>> 1241 E. Catherine St. >>> Ann Arbor MI 48109-5618 >>> 734-615-7826 >>> ********************************************************** >>> Electronic Mail is not secure, may not be read every day, and should >>> not be used for urgent or sensitive issues >>> >>> _______________________________________________ >>> Bioconductor mailing list >>> Bioconductor at r-project.org >>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>> Search the archives: >>> http://news.gmane.org/gmane.science.biology.informatics.conductor >> >> > > > ______________________________________________________________________ > The information in this email is confidential and inte...{{dropped:16}}