Question: understanding siggenes results
0
8.0 years ago by
Assa Yeroslaviz1.4k
Munich, Germany
Assa Yeroslaviz1.4k wrote:
Hallo BioC users, I am using the siggenes package to analyze miRNA-Microarrays. It all run perfectly well and I have got a list of 14 miRNA with significantly differential regulation. Than I wanted to check the values of these miRNA to see the true values (after normalization), just to check the relative expression of these miRNA in comparisons to the rest on the array. The miRNA affy arrays are built differentially. They contain not only drosophila miRNA, but also from other sources. So I did the analysis as always: 1. reading the files: rawData <- ReadAffy() sampleNames(rawData) <- sub("^\\d\\_(.*)\\_\$$miRNA-1\\_0\\_2Xgain\$$.CEL$", "\\1", sampleNames(rawData)) sampleNames(rawData) 2.normalizing using RMA: rma_data <- rma(rawData) 3. filtering the expressionSet only for these probesets from drosophila: dme <- grep("dme", Probe_names) dme_rma <- rma_data[dme,] 4. run the sam analysis with subtset: dme_sam <- sam(dme_rma, cl, var.equal=FALSE, B=100, include.zero=FALSE, gene.names = featureNames(dme_rma), R.fold = 1.5, na.replace=FALSE, rand=123) list.siggenes(dme_sam, 1.7)# gies me back a list of 14 genes. summary(dme_sam,1.7) SAM Analysis for the Two-Class Unpaired Case Assuming Unequal Variances Number of variables having a fold change >= 1.5 or <= 0.6667 : 26 s0 = 0.0514 (The 0 % quantile of the s values.) Number of permutations: 20 (complete permutation) MEAN number of falsely called variables is computed. Delta: 1.7 cutlow: -6.087 cutup: 2.896 p0: 0 Identified Genes: 14 Falsely Called Genes: 0.9 FDR: 0 Identified Genes (using Delta = 1.7): Row d.value stdev rawp q.value R.fold Name 1 132 13.33 0.2369 0.00385 0 14.344 dme-miR-986_st 2 4 7.16 0.0514 0.00769 0 1.666 dme-miR-1001_st 3 126 -6.09 0.0920 0.01154 0 0.546 dme-miR-980_st 4 35 5.19 0.7641 0.01538 0 18.841 dme-miR-193_st 5 133 4.53 0.1673 0.01923 0 1.989 dme-miR-987_st 6 7 4.48 0.0924 0.02308 0 1.563 dme-miR-1004_st 7 100 4.38 0.3418 0.02692 0 3.295 dme-miR-954_st 8 29 4.20 0.2729 0.03077 0 2.571 dme-miR-13a_st 9 37 3.72 0.3097 0.03462 0 2.534 dme-miR-210_st 10 93 3.58 0.2528 0.03846 0 2.126 dme-miR-87_st 11 127 3.57 0.1949 0.04231 0 1.840 dme-miR-981_st 12 3 3.22 0.1529 0.05000 0 1.578 dme-miR-1000_st 13 135 3.00 0.2862 0.05385 0 2.018 dme-miR-989_st 14 118 2.90 0.1537 0.05769 0 1.510 dme-miR-972_st As you can see I have these miRNA with significant deregulation with a delta value of 1.7. Than I looked at the normalized values of some of these genes and the results are very strange. here is a list of three of the genes from the data set: the first one is a downregulated miRNA and the others are the two strongest up-regulated miRNAs. miRNA_id wt1 wt2 wt3 mut1 mut2 mut3 mean_wt man_mut sam results dme-miR-980_st 10.72 10.45 10.48 9.71 9.71 9.61 10.55 9.68 0.55 dme-miR-986_st 4.86 5.43 5.24 9.06 8.71 9.28 5.18 9.02 14.34 dme-miR-193_st 1.75 1.75 2.24 4.69 6.62 7.14 1.91 6.15 18.84 For each line I calculated the mean, just to get an Idea of the two summarized values of each one of them, but this doesn't even get close to the R.Fold values I can see in the sam result table. Does anyone have an explanation for such a behavior? Do I understand it completely wrong? I would appriciate any help you have. Thanks Assa [[alternative HTML version deleted]] mirna affy siggenes • 620 views ADD COMMENTlink modified 8.0 years ago by James W. MacDonald50k • written 8.0 years ago by Assa Yeroslaviz1.4k Answer: understanding siggenes results 0 8.0 years ago by United States James W. MacDonald50k wrote: Hi Assa, On 5/19/2011 7:18 AM, Assa Yeroslaviz wrote: > Hallo BioC users, > > I am using the siggenes package to analyze miRNA-Microarrays. > > It all run perfectly well and I have got a list of 14 miRNA with > significantly differential regulation. > > Than I wanted to check the values of these miRNA to see the true values > (after normalization), just to check the relative expression of these miRNA > in comparisons to the rest on the array. > > The miRNA affy arrays are built differentially. They contain not only > drosophila miRNA, but also from other sources. > So I did the analysis as always: > 1. reading the files: > rawData<- ReadAffy() > sampleNames(rawData)<- sub("^\\d\\_(.*)\\_\$$miRNA-1\\_0\\_2Xgain\$$.CEL$", > "\\1", sampleNames(rawData)) > sampleNames(rawData) > > 2.normalizing using RMA: > rma_data<- rma(rawData) > > 3. filtering the expressionSet only for these probesets from drosophila: > dme<- grep("dme", Probe_names) > dme_rma<- rma_data[dme,] > > 4. run the sam analysis with subtset: > dme_sam<- sam(dme_rma, cl, var.equal=FALSE, B=100, include.zero=FALSE, > gene.names = featureNames(dme_rma), R.fold = 1.5, na.replace=FALSE, > rand=123) > > list.siggenes(dme_sam, 1.7)# gies me back a list of 14 genes. > summary(dme_sam,1.7) > > SAM Analysis for the Two-Class Unpaired Case Assuming Unequal Variances > > Number of variables having a fold change>= 1.5 or<= 0.6667 : 26 > > s0 = 0.0514 (The 0 % quantile of the s values.) > > Number of permutations: 20 (complete permutation) > > MEAN number of falsely called variables is computed. > > Delta: 1.7 > cutlow: -6.087 > cutup: 2.896 > p0: 0 > Identified Genes: 14 > Falsely Called Genes: 0.9 > FDR: 0 > > > Identified Genes (using Delta = 1.7): > > Row d.value stdev rawp q.value R.fold Name > 1 132 13.33 0.2369 0.00385 0 14.344 dme-miR-986_st > 2 4 7.16 0.0514 0.00769 0 1.666 dme-miR-1001_st > 3 126 -6.09 0.0920 0.01154 0 0.546 dme-miR-980_st > 4 35 5.19 0.7641 0.01538 0 18.841 dme-miR-193_st > 5 133 4.53 0.1673 0.01923 0 1.989 dme-miR-987_st > 6 7 4.48 0.0924 0.02308 0 1.563 dme-miR-1004_st > 7 100 4.38 0.3418 0.02692 0 3.295 dme-miR-954_st > 8 29 4.20 0.2729 0.03077 0 2.571 dme-miR-13a_st > 9 37 3.72 0.3097 0.03462 0 2.534 dme-miR-210_st > 10 93 3.58 0.2528 0.03846 0 2.126 dme-miR-87_st > 11 127 3.57 0.1949 0.04231 0 1.840 dme-miR-981_st > 12 3 3.22 0.1529 0.05000 0 1.578 dme-miR-1000_st > 13 135 3.00 0.2862 0.05385 0 2.018 dme-miR-989_st > 14 118 2.90 0.1537 0.05769 0 1.510 dme-miR-972_st > > > As you can see I have these miRNA with significant deregulation with a > delta value of 1.7. > Than I looked at the normalized values of some of these genes and the > results are very strange. > here is a list of three of the genes from the data set: > the first one is a downregulated miRNA and the others are the two strongest > up-regulated miRNAs. > > miRNA_id wt1 wt2 wt3 mut1 mut2 mut3 mean_wt man_mut sam results > dme-miR-980_st 10.72 10.45 10.48 9.71 9.71 9.61 10.55 9.68 0.55 > dme-miR-986_st 4.86 5.43 5.24 9.06 8.71 9.28 5.18 9.02 14.34 > dme-miR-193_st 1.75 > 1.75 2.24 4.69 6.62 7.14 1.91 6.15 18.84 > For each line I calculated the mean, just to get an Idea of the two > summarized values of each one of them, but this doesn't even get close to > the R.Fold values I can see in the sam result table. > > Does anyone have an explanation for such a behavior? > > Do I understand it completely wrong? You are missing the fact that the fold changes that siggenes reports are no longer on the log scale. > 2^(9.68-10.55) [1] 0.5471469 > 2^(9.02-5.18) [1] 14.3204 > 2^(6.15-1.91) [1] 18.89588 My values are slightly off because IIRC, what really happens in siggenes (and for sure happens in SAM) is that the data are 'unlogged' first, then the mean is computed, then the ratio. Best, Jim > > I would appriciate any help you have. > > Thanks > > Assa > > [[alternative HTML version deleted]] > > _______________________________________________ > 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

Content
Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.