Search
Question: DEXSeqHTML Error on condition variable
0
10 months ago by
mittjohns0 wrote:

I used DEXSeq package to analyze RNA-seq data with two genotypes: KO vs WT, and two treatments: treated vs untreated. I are interested in the interaction between genotype and treatment, i.e. the effects of KO + treated. I preprocessed the data in R using summarizeOverlaps function followed by by DEXSeqDataSetFromSE. I used the following code chunk. Everything worked fine till the last reporting step with DEXSeqHTML. It asked for a column called condition in the sample annotation of DEXSeqDataSet, which I don’t use throughput all analysis. Note that I did get 21 significant exon events as shown in the end.

I used DEXSeq_1.22.0. let me know if the sessionInfo is needed. It looks like DEXSeqHTML function need to be updated. Please help or let me know what can I do? Thank you.

My code:

colData(se)$Genotype = samples$Genotype

colData(se)$Treatment = samples$Treatment

library("DEXSeq")

dxd = DEXSeqDataSetFromSE(se,

design= ~ sample + exon + Genotype:exon + Treatment:exon + Genotype:Treatment:exon)

dxd = estimateSizeFactors(dxd)

dxd = estimateDispersions(dxd)

dxd = testForDEU(dxd, reducedModel = ~sample + exon + Genotype:exon + Treatment:exon)

dxr1 = DEXSeqResults(dxd)

colors <- brewer.pal(4, "Set2")

DEXSeqHTML(dxr1, FDR=0.1 , color = colors)

Output:

>table ( dxr1\$padj < 0.1 )

FALSE   TRUE

214747     21

> DEXSeqHTML(dxr1, FDR=0.1 , color = colors)

Error in DEXSeqHTML(dxr1, FDR = 0.1, color = colors) :

The value of the parameter fitExpToVar, 'condition', is not a variable from the annotation of the samples. Please specify a column name of the colData slot of the DEXSeqDataSet object.

modified 10 months ago by Alejandro Reyes1.5k • written 10 months ago by mittjohns0
0
10 months ago by
Alejandro Reyes1.5k
Dana-Farber Cancer Institute, Boston, USA
Alejandro Reyes1.5k wrote:

Hi mittjohns,

The parameter fitExpToVar of the function DEXSeqHTML requires a variable of the experiment that you are analyzing (it has the value "condition" by default but in your dataset that could be either "Genotype" or "Treatment").  However, this won't be useful to visualize the interaction effect between Genotype and Treatment on exon usage since visualizing these kind of complex interactions in the DEXSeq plot is currently not supported. One option could be to combine the Genotype and Treatment columns into a single column and specify the name of this column in fitExpToVar. Another would be to just look at the plot of normalized counts in each sample.

Alejandro