I'm looking for a filtering method that it does't use genewise variances
Hi guys
I have a dataset with 100 sample and 54000 feature,I used to think that if I filter those of probIDs with little variation, I can reduce the number of features and then in multiple testing with limma it gives me fewer errors and finally fewer noises, but now I found that if I use genefilter package of bioconductor, It changes the the distribution of variance and will interfere with limma package
Let me know if I am mistaken and tell me if there are other options except genefiltering package
TNX
• 2,405 views
•
link
updated
by
Ram
•
written
by
Shamim Sarhadi
•
written
by
Shamim Sarhadi
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Microarray analysis error
written by biology_inform •
-
How to simulate phenotype from two random multivariant normal distributed variable and calculate st…
written by gaaraqrq •
-
Problem in hierarchical clustering
written by Calangoa •
-
Is right normalize metabolomics "counts" by Variance Stabilized Transformation using VST DESeq2 pac…
written by panconchoclo •
-
which set of results are more reliable and precise for DEG analysis, limma vs GSEA ?
written by Shamim Sarhadi
-
Selecting present prob-id after performing pa.calls function from panp package on eSet
written by Shamim Sarhadi
-
Automation of R package loading
written by hakimelakhrass
-
Adding a variable as a varlabel to my dataset
written by Shamim Sarhadi
-
Accurate estimation of variance of ratios of microarray probe signals
written by C0mmanderKeen •
-
No Go Terms For Genes When Using The Gostat Package In Bioconductor
written by shaikhfarahdeeba •
Hint: filter after the testing, just make sure to not filter on something linked to the test.
But I want to filter non-informative probID before the test(those with little variation or those with not express in most of samples ) but now the problem is that the limma in it's vintage says that some filtering method have conflict with limma as I said before !!
Why are you fixated on filtering before testing? You can just readjust the p-values after filtering post-testing and this is the recommended method. If you just did what I suggested then limma would work fine and you would get good results.
BTW, do make sure that your p-values are uniformly distributed over variance if you're going to filter by that. Whatever you filter by MUST not correlate with the test or you're p-hacking.
Thank you Devon
I do not insist on filtering before testing !!! I just thought it gives me a better result, It seems your suggestion is more reasonable
tnx again