normalization combination of noob + BMIQ/Quantile
Entering edit mode
Lin ▴ 50
Last seen 3.2 years ago

Hi all,

I use minfi for pre-processing of EPIC methylation data, so from my last step I have an RGSet. Now I've read that it is good to combine noob with BMIQ or preprocessQuantile for normalization.

However, I wonder how this can be done in R with the different output objects. Assume I do noob first (noob= preprocessNoob(RGSet), the output is a GenomicRatioSet I guess.

How can I then do preprocess Quantile (needs RGSet) or BMIQ (needs beta matrix?) afterwards?

Thanks for help.

minfi bmiq noob preprocessQuantile champ • 2.3k views
Entering edit mode
Last seen 9 minutes ago
United States

Quantile normalization doesn't need an RGSet. Why do you think that? From ?preprocessQuantile:


     preprocessQuantile(object, fixOutliers = TRUE, removeBadSamples = FALSE,
                        badSampleCutoff = 10.5, quantileNormalize = TRUE,
                        stratified = TRUE, mergeManifest = FALSE, sex = NULL,
                        verbose = TRUE)


  object: An object of class 'RGChannelSet' or '[Genomic]MethylSet'.

And preprocessFunnorm by default first calls preprocessNoob, and then does an arguably better/smarter quantile normalization using the first N PCs of the background probes. So you could just use that. In addition, ?BMIQ (from wateRmelon) says


     BMIQ(beta.v, design.v, nL = 3, doH = TRUE, nfit = 50000, th1.v = c(0.2, 0.75), th2.v = NULL, niter = 5, tol = 0.001, plots = TRUE, sampleID = 1, pri=TRUE)
     ## S4 method for signature 'MethyLumiSet'
     BMIQ(beta.v, nL=3, doH=TRUE, nfit=5000, th1.v=c(0.2,0.75), th2.v=NULL, niter=5, tol=0.001, plots=FALSE,  pri=FALSE ) 
     CheckBMIQ(beta.v, design.v, pnbeta.v)


  beta.v: vector consisting of beta-values for a given sample, or a
          MethyLumiSet or MethylSet containing multiple samples. For
          the MethyLumiSet and MethylSet methods, this is the only
          required argument, and the function will be run on each

So you don't need a beta matrix for BMIQ either.

Entering edit mode

Thanks James, so I can first use noob, then I get a MethylSet, then I can use preprocessQuantile or BMIQ. I guess I was also a bit confused because especially for BMIQ there are many different packages.

However, one more question: Normalization is always an issue with all the different possibilities. I decided to use Quantile or BMIQ and not Funnorm, because I have samples from the same tissue and also not sth like cancer-control. Isn't it the case that then Quantile is better than Funnorm (I refer to the study of Fortin et al. for example, or Liu& Sigmund who found that noob+BMIQ perform good)?

Entering edit mode

It's up to you which normalization you think is better. I wouldn't recommend using a support site to decide how best to analyze your data.


Login before adding your answer.

Traffic: 943 users visited in the last hour
Help About
Access RSS

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

Powered by the version 2.3.6