I am using minfi to analysis idat files, in minfi introduction there is a function that retrieves the out of band probes but this function only retrieves. i dont know how can i remove these probes. Can anyone to introduce a function for removing out of band probes?
You don't need to remove them as such, it's just a convenience function. Once you're created your RedGreenChannelSet you can use preprocessNOOB() or preprocessFunnorm() to normalise your data which will fix most issues since both functions allow you to carry out background correction. Check out the vignette for information on the normalisation functions. Both functions conveniently output a GenomicMethylRatioSet which you can then use getBeta() on to get beta-values, or getM() to get M-values matrices.
If you want additional QC assurance you can run the qcReport() function on your RedGreenChannelSet before you normalise to look for problem probes. You can also use detectionP() to find probes which fail detection threshold and remove samples if they have a sufficient number of probes missing. 1% - 5% is usually a good threshold.
Additionally, once you have generated your GenomicMethylRatioSet and have a beta-values matrix, you can run the function rmSNPandCH() from the package DMRcate to remove various problem probes, noisy sex chromosomes etc.
Thanks, does preprocessNOOB() or preprocessFunnorm() remove beadcounts, too? i want to remove beadcounts <3. It sounds there is not any function to remove this.
Normalising doesn't 'remove' any probes. It just adjusts the intensities based on some measurements - for NOOB it uses the out of band probes, and for Funnorm it uses NOOB and also adjusts for variation by calculating variation from the control arrays and removing variation corresponding to n principal components.
There are ways of doing this but I haven't really found the need.
As, human methylation 450k pipline shows one of pre-processing steps is removing probes with bead-count <3 but why you think we dont need removing.