CummeRbund csDendro error: "need finite ylim values" when replicates=TRUE

[mebbert]

All,

I was trying to create a dendrogram (including replicates) in cummeRbund using the csDendro function, but I get an error stating: "Error in plot.window(...) : need finite 'ylim' values". I'm using the following command:

cuff<-readCufflinks()
dend.rep<-csDendro(genes(cuff),replicates=T)

Does anyone know why?

Useful information:

• I'm using R 2.15.1, tophat2 2.0.3, cufflinks 2.0.2, and cummeRbund 1.99.2
• I have been following the protocol from Nature Protocols (http://www.nature.com/nprot/journal/....2012.016.html), but I'm getting a lot of NOTEST results (49,956) from cuffdiff. I'm currently awaiting results from two test runs testing '-c 0' and '--min-outlier-p 0' to see if that helps with the NOTEST results and to see if it's related to the above cummeRbund error. I'm testing these parameters as per Cole Trapnell's suggestions (http://seqanswers.com/forums/showthr...merbund&page=2)
• I have two conditions with 30 and 23 replicates (mostly biological but a few technical)

Thanks for your help!

[lgoff]

Hi mebbert,

Can you tell me if the output of:

repFpkmMatrix(genes(cuff))

contains any missing values? I'm trying to figure out if this is the problem. Several users have had this problem.

Thanks in advance,

Cheers,
Loyal

[mebbert]

Thanks for the response, Loyal. In short, yes there are missing values. Here's partial output from:

x<-repFpkmMatrix(genes(cuff))
head(x)

You will see that XLOC_000006 has only NA values. When looking at the first 30 rows there are two such situations.

            FFPE_0 FFPE_1 FFPE_2 FFPE_3 FFPE_4 FFPE_5 FFPE_6 FFPE_7    FFPE_8 FFPE_9 FFPE_10 FFPE_11 FFPE_12 FFPE_13 FFPE_14 FFPE_15 FFPE_16 FFPE_17 FFPE_18 FFPE_19 FFPE_20 FFPE_21 FFPE_22 FFPE_23
XLOC_000001      0      0      0      0      0      0      0      0 0.0000000      0       0       0       0       0       0       0       0       0       0       0       0       0       0       0
XLOC_000002      0      0      0      0      0      0      0      0 0.0000000      0       0       0       0       0       0       0       0       0       0       0       0       0       0       0
XLOC_000003      0      0      0      0      0      0      0      0 0.0000000      0       0       0       0       0       0       0       0       0       0       0       0       0       0       0
XLOC_000004      0      0      0      0      0      0      0      0 0.0000000      0       0       0       0       0       0       0       0       0       0       0       0       0       0       0
XLOC_000005      0      0      0      0      0      0      0      0 0.0726539      0       0       0       0       0       0       0       0       0       0       0       0       0       0       0
XLOC_000006     NA     NA     NA     NA     NA     NA     NA     NA        NA     NA      NA      NA      NA      NA      NA      NA      NA      NA      NA      NA      NA      NA      NA      NA

[lgoff]

Hi Mebbert, that's what I figured. This is something that will have to be fixed in the next release of cummeRbund and cuffdiff. Can you confirm that you are using cufflinks v2.0.2 as well?

Thanks,
Loyal

[mebbert]

Yes, I am using Cufflinks 2.0.2. Do you know if there is an estimated release date? Could this be related to the high number of NOTESTS that I'm seeing from CuffDiff?

I should point out that I get the following warnings when I read in the data with readCufflinks():

Warning messages:
1: Removed 32139 rows containing missing values (geom_point).
2: Removed 30203 rows containing missing values (geom_point).
3: Removed 37458 rows containing missing values (geom_point).

Furthermore, as per Cole Trapnell's advice (HERE), I reran the data with '-c 0' and '--min-outlier-p 0' in two separate analyses. NOTESTS decreased to 32137 with the '-c 0' option and did not change with '--min-outlier-p 0'.

Any ideas?

Thanks!