[BioC] Difference in DABG between XPS package and APT

[cstrato]

Dear Zaid,

The differences may be caused as follows:

- Although you did not mention how you have created the root scheme file 
I assume that you use the newest Affymetrix annotation files version 
na32. Please note that these annotation files use human genome build 
hg19, whereas the file HuEx-1_0-st-v2.r2.dt1.hg18.full.mps uses hg18.

- The metaprobeset file HuEx-1_0-st-v2.r2.dt1.hg18.full.mps that you are 
using is not only based on hg18, but was created on Oct 2006, and is 
thus completely outdated. To be able to compare APT with XPS you need to 
create a new metaprobeset file, e.g. using xps function metaProbesets().

- The clf-file you are using, i.e. HuEx-1_0-st-v2.r2.crosshyb_1.na24.clf 
is not the official Affymetrix clf-file, which is called 
HuEx-1_0-st-v2.r2.clf

- If you want to lower the number of PercentPresent you can adjust 
alpha1 and alpha2 to get usable P/M/A calls for transcripts.


Please have a look at vignette APTvsXPS.pdf, especially chapter 4.3 
where I have compared DABG calls from APT vs XPS. Furthermore, please 
look at the corresponding script xps/examples/script4xps2apt.R which 
contains the complete source code used in vignette APTvsXPS.

Best regards,
Christian


On 8/31/12 7:06 PM, zaid haddad [guest] wrote:
> Hello,
>
> I have a question regarding XPS and APT when computing DABG option.
>
> Here's my XPS command:
>
> #begin XPS code
> dabg.call(root.data, filedir = root.path, filename = paste("tmpdt", "RMA", root.filename, sep="_"), option = "probeset", exonlevel = "all", verbose = TRUE)
>
> #end
>
> Here's my APT command line:
>
> #begin
> apt-probeset-summarize -a dabg -feat-details -b /Data/affx.tech.files/HuEx-1_0-st-v2.r2/HuEx-1_0-st-v2.r2.antigenomic.bgp -p /Data/affx.tech.files/HuEx-1_0-st-v2.r2/HuEx-1_0-st-v2.r2.pgf -c /Data/affx.tech.files/HuEx-1_0-st-v2.r2/HuEx-1_0-st-v2.r2.crosshyb_1.na24.clf -s /Data/affx.tech.files/HuEx-1_0-st-v2.r2/HuEx-1_0-st-v2.r2.dt1.hg18.full.mps --qc-probesets /Data/affx.tech.files/HuEx-1_0-st-v2.r2/HuEx-1_0-st-v2.r2.qcc -o dabg/ --subsample-report --cel-files cel_list.txt
>
> #end
>
> I get different p-values and different Percent Present calls. (Highly correlated though).
>
> #R code
>   cor(apt.detcal$all_probeset_percent_called,xps.detcal$PercentPresent)
> #[1] 0.9978199
>
> #difference in PSRs between XPS and APT dabg output
>   length(setdiff(xps.dabg[,2], rownames(apt.dabg)))
> #[1] 15332
> #difference in PSRs between APT and XPS dabg output
>   length(setdiff( rownames(apt.dabg), xps.dabg[,2]))
> #[1] 2904
> #Total # of PSRs that differ
>   2904+15332
> #[1] 18236
> #PSRs in common
>   length(intersect(xps.dabg[,2], rownames(apt.dabg)))
> #[1] 1381327
>
> The APT output is always lower in the PercentPresent.
>
> Thanks in Advance
>
>   -- output of sessionInfo():
>
> R version 2.13.1 (2011-07-08)
> Platform: x86_64-pc-linux-gnu (64-bit)
>
> locale:
>   [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
>   [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
>   [5] LC_MONETARY=C              LC_MESSAGES=en_US.UTF-8
>   [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
>   [9] LC_ADDRESS=C               LC_TELEPHONE=C
> [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
>
> attached base packages:
> [1] tools     splines   stats     graphics  grDevices utils     datasets
> [8] methods   base
>
> other attached packages:
>   [1] xps_1.12.1          abind_1.4-0       survival_2.36-14
>   [4] affxparser_1.24.0   fastmatch_1.0-4     glmnet_1.7.1
>   [7] Matrix_0.999375-50  earth_3.2-3         plotrix_3.4-3
> [10] plotmo_1.3-2        leaps_2.9           rpart_3.1-50
> [13] HDclassif_1.2.1     MASS_7.3-13         randomForest_4.6-6
> [16] e1071_1.6           class_7.3-3         oligo_1.14.0
> [19] oligoClasses_1.14.0 frma_1.3.10         Biobase_2.12.2
> [22] caret_5.15-023      cluster_1.14.0      reshape_0.8.4
> [25] plyr_1.7.1          lattice_0.19-30     matrixStats_0.4.3
> [28] R.methodsS3_1.4.2   abind_1.4-0         foreach_1.4.0
> [31] itertools_0.1-1     iterators_1.0.5
>
> loaded via a namespace (and not attached):
> [1] affy_1.30.0           affyio_1.20.0         Biostrings_2.20.4
> [4] codetools_0.2-8       DBI_0.2-5             grid_2.13.1
> [7] IRanges_1.10.6        preprocessCore_1.14.0
>
>
> --
> Sent via the guest posting facility at bioconductor.org.
>