[BioC] Understanding "minfi" package & its errors: for the analysis of 450K Methylation chip
Kasper Daniel Hansen
kasperdanielhansen at gmail.com
Tue Sep 4 15:02:31 CEST 2012
On Tue, Sep 4, 2012 at 4:09 AM, pooja mandaviya
<pooja.mandaviya at gmail.com> wrote:
> Hi all,
> I am kind of new to using R and I have recently started working with
> Illumina 450K methylation chip and am trying to use the package "minfi" for
> my 780 sample data. While I run my data, I was stuck at small errors and
> doubts which I wanted to ask and request for your help. I here list them
> 1) densityPlot (RGset, sampGroups = pd$Sample_Group, main = "Beta", xlab =
> For this, i get an error saying,
> Error in density.default(newX[;i],...):
> need at least 2 points to select a bandwidth automatically.
> I get the above error most of the time with most of the datasets i run.
> Could you help me know how do I get rid with it?
This looks weird, given that I would expect your RGset to have
100,000's of rows. What does
> 2)Mset.swan <- preprocessSWAN(RGsetEx, MsetEx)
> While running this command, it always says,
> Normalizing array 1 of 6
> Normalizing array 2 of 6
> & so on..
> My question here is that it always shows this and always normalizes the
> first 6 arrays/samples, but how about the other datasets having more than 6
> samples? I sometimes run a dataset of 100-150 samples. It still shows
> normalizing 6 samples. How about the rest of the samples?
Well, I hope you use your own data instead of RGsetEx, MsetEx. The i
out of n message uses the number of columns (samples) of the input
> 3) Most of the test datasets run fine, apart from small errors which i
> listed above. I am also able to get the final plots. However, my main
> dataset which i have to work on, contains 780 samples. I have tried running
> this through minfi as well. But I always get errors running through most of
> the commands with it. Just to list them here; For commands like QCReport,
> densityBeanPlot, controlStripPlot, MSet.norm getBetam getM & mdsPlot, I get
> similar errors like
>> Error: cannot allocate vetor of size 3.6 Gb.
>> Warning: BISULFITE Conversion 2 probes outside plot range.
> So my question is that, is minfi not able to support very large datasets?
> (In my case: 780 samples)
You either need more RAM or you need to run a 64bit R, perhaps both.
> 4) Last question: While normalizing, does minfi also take care of all the
> batch effects?
No, in general you do not remove batch effects by normalization.
> Best Wishes,
> Pooja Mandaviya
> Department of Clinical Chemistry
> Erasmus Medical Center, Rotterdam
> The Netherlands
> p.mandaviya at erasmusmc.nl
More information about the Bioconductor