Question

How to remove spike ins from illumina beadchip data

1

Entering edit mode

4.4 years ago

maria2019 ▴ 250

Hi,

I am very new to microarray analysis. I have some cancer and control samples (idat files) from illumina beadchip to analyze. I was following the tutorial on limma from ( https://www.bioconductor.org/packages/release/bioc/vignettes/limma/inst/doc/usersguide.pdf ), page 107.

When I check the EListRaw object, I get lots of ERCC as well which I did not expect!! (code and results are below).

Why do I have spike ins?
How can I remove them? (should I actually remove them?)

I ignored this and went through the downstream analysis but the FDR that I get is so hight (0.5 and above) which I think might be the result of I not removing ERCCs.

    $ idatfiles = dir("path", pattern = "idat",full.names = TRUE)
    $ bgxfile <- "my.bgx"
    $ x = read.idat(idatfiles, bgxfile)
    $ x$other$Detection <- detectionPValues(x)
    $ table(x$genes$Status)

>               biotin            cy3_hyb      ERCC-00002-02      ERCC-00003-01 
                         2                  6                  1                  1 
             ERCC-00004-01      ERCC-00009-01      ERCC-00012-01      ERCC-00013-01 
                         1                  1                  1                  1 
             ERCC-00014-02      ERCC-00016-01      ERCC-00017-02      ERCC-00019-01 
                         1                  1                  1                  1 
             ERCC-00022-02      ERCC-00024-02      ERCC-00025-01      ERCC-00028-02 
                         1                  1                  1                  1 
             ERCC-00031-02      ERCC-00033-01      ERCC-00034-02      ERCC-00035-02 
                         1                  1                  1                  1 
             ERCC-00039-01      ERCC-00040-01      ERCC-00041-01      ERCC-00042-01 
                         1                  1                  1                  1 
             ERCC-00043-01      ERCC-00044-02      ERCC-00046-01      ERCC-00048-01 
                         1                  1                  1                  1 
             ERCC-00051-01      ERCC-00053-01      ERCC-00054-01      ERCC-00057-01 
                         1                  1                  1                  1 
             ERCC-00058-02      ERCC-00059-01      ERCC-00060-01      ERCC-00061-02 
                         1                  1                  1                  1 
             ERCC-00062-01      ERCC-00067-02      ERCC-00069-02      ERCC-00071-01 
                         1                  1                  1                  1 
             ERCC-00073-01      ERCC-00074-01      ERCC-00075-01      ERCC-00076-02 
                         1                  1                  1                  1 
             ERCC-00077-01      ERCC-00078-01      ERCC-00079-01      ERCC-00081-02 
                         1                  1                  1                  1 
             ERCC-00083-01      ERCC-00084-01      ERCC-00085-01      ERCC-00086-01 
                         1                  1                  1                  1 
             ERCC-00092-02      ERCC-00095-01      ERCC-00096-02      ERCC-00097-01 
                         1                  1                  1                  1 
             ERCC-00098-02      ERCC-00099-01      ERCC-00104-01      ERCC-00108-02 
                         1                  1                  1                  1 
             ERCC-00109-02      ERCC-00111-01      ERCC-00112-02      ERCC-00113-01 
                         1                  1                  1                  1 
             ERCC-00116-02      ERCC-00117-02      ERCC-00120-01      ERCC-00123-01 
                         1                  1                  1                  1 
             ERCC-00126-02      ERCC-00130-01      ERCC-00131-02      ERCC-00134-01 
                         1                  1                  1                  1 
             ERCC-00136-01      ERCC-00137-02      ERCC-00138-01      ERCC-00142-02 
                         1                  1                  1                  1 
             ERCC-00143-01      ERCC-00144-02      ERCC-00145-01      ERCC-00147-01 
                         1                  1                  1                  1 
             ERCC-00148-01      ERCC-00150-01      ERCC-00154-02      ERCC-00156-01 
                         1                  1                  1                  1 
             ERCC-00157-02      ERCC-00158-01      ERCC-00160-02      ERCC-00162-01 
                         1                  1                  1                  1 
             ERCC-00163-01      ERCC-00164-01      ERCC-00165-01      ERCC-00168-01 
                         1                  1                  1                  1 
             ERCC-00170-01      ERCC-00171-01       housekeeping           labeling 
                         1                  1                  7                  2 
        low_stringency_hyb           negative            regular 
                         8                770              47231

microarray limma beadchip illumina ERCC • 1.1k views

ADD COMMENT • link updated 4.4 years ago by Kevin Blighe 88k • written 4.4 years ago by maria2019 ▴ 250

score 2 · Accepted Answer · 2019-12-25

You need to leave the control probes in the data for the purpose of background correction and normalisation. Then, if you perform background correction and normalisation via neqc(), these control probes should be automatically removed.

After normalisation, you can do further filtering based on the detection p-values. Any other control probes that still remain in the data may be identified via x$genes$Source. Others that can be filtered out include those with no gene symbol (x$genes$Symbol == "")

Thus, filtering that I perform post-normalisation is like this:

Control <- project.bgcorrect.norm$genes$Source=="ILMN_Controls"
NoSymbol <- project.bgcorrect.norm$genes$Symbol == ""
isexpr <- rowSums(project.bgcorrect.norm$other$Detection <= 0.05) >= 3

project.bgcorrect.norm.filt <- project.bgcorrect.norm[!Control & !NoSymbol & isexpr, ]

dim(project.bgcorrect.norm)
dim(project.bgcorrect.norm.filt)

Kevin