Entering edit mode
BJ Chen
▴
60
@bj-chen-6670
Last seen 10.3 years ago
Hi,
I am trying to run DESeq2 analysis on sample design like this:
sample replicate treatment
A 1 no
A 2 no
A 1 yes
A 2 yes
B 1 no
B 2 no
B 1 yes
B 2 yes
(repeated for 4 different samples. eg. A-D).
The interests of DE effect include treatment on each sample and
treatment
over all.
I have searched online and found previously suggested model as
~ treatment + sample:replicate + sample:treatment.
However, when I called DESeqDataSetFromMatrix(readcount, sampleinfo,
~treatment+sample:replicate+sample:treatment), it first complained the
matrix is not full rank.
I tried ignoreRank option, but then I got error when I called DESeq()
(default parameters):
error: inv(): matrix appears to be singular
Error in eval(expr,
envir,
enclos) : inv(): matrix appears to be singular
In addition: Warning message:
In
fitNbinomGLMs(objectNZ[fitidx, , drop = FALSE], alpha_hat =
alpha_hat[fitidx], : 25rows had non-positive estimates
of
variance for coefficients, likely due to rank deficient model matrices
without betaPrior
If I exclude the replicate ( eg.
design=~treatment+sample+sample:treatment), it run through without
errors.
However, I would like to take into account the replicates, as they are
paired samples for the treatment.
I will appreciate any help/suggestions.
Thanks,
BJ
Session info is included in the bottom.
R
versio(2014-04-10)4-04-10)
Platform: x86_64-unknown-linux-gnu (64-bit)
attached base packages:
[1] parallel stats graphics grDevices utils datasets
methods
[8] base
other attached
packages:
[1] DESeq2_1.4.5 RcppArmadillo_0.4.200.0
Rcpp_0.11.1
[4] GenomicRanges_1.14.4 XVector_0.2.0
IRanges_1.20.7
[7] BiocGenerics_0.8.0
loaded via a namespace (and not
attached):
[1] annotate_1.40.1 AnnotationDbi_1.24.0
Biobase_2.22.0
[4] DBI_0.2-7 genefilter_1.44.0
geneplotter_1.40.0
[7] grid_3.1.0 lattice_0.20-29 locfit_1.5-9.1
[10] RColorBrewer_1.0-5
RSQLite_0.11.4 splines_3.1.0
[13] stats4_3.1.0 survival_2.37-7 XML_3.98-1.1
[16] xtable_1.7-3
[[alternative HTML version deleted]]