[BioC] DiffBind error with DBA_DESEQ_GLM and warning with DBA_EDGE_GLM

Elisabetta Manduchi manduchi at pcbi.upenn.edu
Mon Jul 16 20:40:12 CEST 2012



Hello,
I'm using DiffBind version 1.2.0 with R 2.15.0 on linux.
I've created a dba object data.deseq and run:

> data.deseq =
dba.count(data.deseq,score=DBA_SCORE_TMM_READS_EFFECTIVE,bRemoveDuplicates=T,bCorPlot=F)
> data.deseq =
dba.contrast(data.deseq,categories=DBA_TISSUE,block=DBA_FACTOR)

So data.deseq looks like this:

24 Samples, 212418 sites in matrix:
     ID     Tissue Factor Condition Replicate Intervals
1   A2       Arch      2         M         1    212418
2   A3       Arch      3         F         2    212418
3   A4       Arch      4         M         3    212418
4   A5       Arch      5         F         4    212418
5   A7       Arch      7         F         5    212418
6   D2 Descending      2         M         1    212418
7   D3 Descending      3         F         2    212418
8   D4 Descending      4         M         3    212418
9   D5 Descending      5         F         4    212418
10  D7 Descending      7         F         5    212418
11 A10       Arch     10         F         6    212418
12 A11       Arch     11         M         7    212418
13 A12       Arch     12         M         8    212418
14 A13       Arch     13         M         9    212418
15 A14       Arch     14         M        10    212418
16 A15       Arch     15         F        11    212418
17 A17       Arch     17         F        12    212418
18 D10 Descending     10         F         6    212418
19 D11 Descending     11         M         7    212418
20 D12 Descending     12         M         8    212418
21 D13 Descending     13         M         9    212418
22 D14 Descending     14         M        10    212418
23 D15 Descending     15         F        11    212418
24 D17 Descending     17         F        12    212418

1 Contrast:
   Group1 Members1     Group2 Members2 Block1Val InBlock1 Block2Val InBlock2
1   Arch       12 Descending       12        10        2 11        2
   Block3Val InBlock3 Block4Val InBlock4 Block5Val InBlock5 Block6Val
InBlock6
1        12        2        13        2        14        2 15        2
   Block7Val InBlock7 Block8Val InBlock8 Block9Val InBlock9 Block10Val
InBlock10
1        17        2         2        2         3        2 4         2
   Block11Val InBlock11 Block12Val InBlock12
1          5         2          7         2


1. When I then try to run:
data.deseq =
dba.analyze(data.deseq,method=DBA_DESEQ_GLM,bSubControl=F,bCorPlot=F)

I get the following error:
Error in .local(object, ...) :
   None of your conditions is replicated. Use method='blind' to estimate
across conditions, or 'pooled-CR', if you have crossed factors.
In addition: Warning message:
Unable to use tagwise dispersion estimates with GLM

I'm not clear why it's not seeing the replicates, since I have 12 for
each of Arch and Descending. Am I setting things up incorrectly?

2. Note that I've also analyzed the same dba with method DBA_EDGER_GLM. This
ran, but gave me the following warning on which I'm unclear

Warning message:
In maximizeInterpolant(spline.pts, apl.smooth[j, ]) : Divergence

and it only found 154 significant intervals out of the over 212000 input
sites, which seems too low.

Anybody can advice on the above? Thanks,
Elisabetta



More information about the Bioconductor mailing list