I have the idat file form a same sample in 450k and EPIC.
What is the best way to compare those results?
Can i compare raw data or should I apply any specific normalization before comparison?
•
written
by
Tanvir Ahamed
We have now tested multiple functionalities of the minfi package for the analysis of EPIC arrays. We have performed a comparison of the different normalization methods in minfi, and the results are presented in our recent preprint: http://biorxiv.org/content/early/2016/07/23/065490
Hope this helps,
Jean-Philippe
You would apply some sort of normalization.
Some some normalization strategies may work better for 450k vs EPIC arrays (for example, the proportion of Type I vs Type II probes is different), but I've found that the Illumina normalization seems to work pretty well in both cases.
You can apply Illumina normalization within Genome Studio, or the minfi Bioconductor package using preprocessIllumina(). This applies a background adjustment and normalization to controls.
It would also be best to have multiple samples to compare differential methylation (at the site or region level). For example, if you apply a delta beta threshold of 0.2, that might be greater than the difference in beta values obtained from different normalization methods (although it could affect marginally significant candidates).
Thanks for your suggestion .
Can you please tell me, it it possible to get the same result (equivalent value) for intensity on common probe for 450k and EPIC. So for that, can i compare raw data or any further normalization or other procedure if required ?
I believe there are more options for data type export for GenomeStudio (which includes intensity values, as well as other metrics).
Even though the probes are different, you may notice some subtle differences. For example, you may observe scores/predictions that are correlated with at trait on EPIC arrays (but with greater absolute differences than you might expect from the 450k array).
You may also find that some probes are more likely to be missing across samples than others (however, that may not be as directly relevant to your question about the same probes on the 450k-versus-EPIC array). I'm sure there must be at least one study with the same sample processed with the same array. However, each group can define a different set of "possibly problematic" probes, so I want to be careful about generally saying something can't be used.
So, if you have something specific to test, I would probably do the following:
1) Check that the probes are in fact present on the EPIC array
2) If all probes in question are present, see if you can find independently cohorts with similar measurements (but with different arrays). You can then test the 450k array assignment method against new EPIC cohorts.
I also apologize for the delay (I just noticed that this was from ~3 years ago). I think I got a notification of an answer that might have been deleted after it was posted (but I saw this question under a previous answer that I provided).
The minfi package provides tools for analyzing Illumina's Methylation arrays, specifically the 450k and EPIC (also known as the 850k) arrays.Arrays offer a reliable cost effective way of analysing multiple targets on a large scale. The Genome Centre has offered Illumina methylation arrays since their original golden gate assay. Every student will get any type of academic paper and writing tips from the different resources and make his/her educational life run more smoothly and effectively.
I have found many differences between results from EPIC and 450k even after performing illumina normalization. Has anyone else come across this/have any ideas how to deal with this?
Check this paper from minfi developers: https://doi.org/10.1093/bioinformatics/btw691
We discuss methods for the joint analysis and normalization of data from the HumanMethylation450 (‘450k’) and EPIC platforms.
Log in to answer this question.
