[BioC] error when estimating Dispersions for count table data in DESeq

[Simon Anders]

Hi

On 08/07/14 07:45, 梁芳 wrote:
> [...] command "cds = estimateDispersions( cds )". The error message was:
>
> Error in if (nr < 2) stop("nrow(modelMatrix) must be >=2.") :
>    argument is of length zero


> Following is the code:
> library("DESeq");
> CountTable = read.table("test.txt", header=TRUE, row.names=1);
> Design = data.frame(row.names = colnames(CountTable), condition = c("treated", "treated", "untreated", "untreated"));
> cds = newCountDataSet(CountTable, Design);
> cds = estimateSizeFactors(cds);
> cds = estimateDispersions(cds);

For simple two-group comparisons, you should pass a factor rather than a 
data frame. Otherwise, you have to specify explicitly when calling 
'estimateDispersion' which column of the data frame to use.

Try:
   cds = newCountDataSet(CountTable, Design$condition)
instead of
   cds = newCountDataSet(CountTable)

and it should work.

HOWEVER: Please consider starting over and using DESeq2, the successor 
to DESeq. We have made considerable improvements there, compared to 
DESeq, both in terms of usability (your problem would not have arisen 
with DESeq2) and of inferential power.

   Simon