Error in running "diffExpressedVariants" function in kissDE R package

Question

[c.evang]

Dear all, I want to run kissDE R package on my set of SNPs, that I found using KisSplice software. I'm using this code:

snpB<-kissplice2counts("results_all_reads_mDr_all_reads_WW_k25_coherents_type_0a.fa", pairedEnd=FALSE)
conditions<-c("mDr","WW")
resB<-diffExpressedVariants(snpB, conditions, pvalue=1)

But it gives me the following error:

"Error in rowMeans(Y2[, idxA]): 'x' must be an array of at least two dimensions\n An error occured, unable to fit models on data".

Any suggestions will be appreciated.

Thanks

Answer 1

[vincent.lacroix]

Dear c.evang,

KissDE requires replicates. This is most likely the cause of your problem. I undertand you have two biological conditions (mDr, and WW). If you have replicates for each condition, then you need to run KisSplice on all replicates of all conditions. Those are 4 reads files (or possibly 8 reads file if you have paired-end reads). The command line should look like this

kissplice -r mDr1.fa -r mDr2.fa -r WW1.fa -r WW2.fa

this will produce an output file such as: results_mDr1_mDr2_WW1_WW2_type0.fa you can then run KissDE:

snpB<-kissplice2counts("results_mDr1_mDr2_WW1_WW2_type0.fa", pairedEnd=FALSE)
conditions<-c("mDr","mDr","WW","WW")
resB<-diffExpressedVariants(snpB, conditions, pvalue=1)

If you do not have replicates, then you cannot use KissDE.

[c.evang]

Thank you very much vincent.lacroix.

So I will repeat kisSplice using replicates and then kissDE.

Thanks again Chiara

[c.evang]

Dear Vincent.lacroix,

I tried running kissplice using replicates, but another error occured while running kissDE:

[1] "Error in paste(pathToFigs, \"/models.png\", sep = \"\"): object 'pathToFigs' not found\n An error occured, unable to fit models on data."

Any suggestions?

Thanks

[jfo]

Hi @vincent.lacroix,

I also encountered an almost similar error:

"Error in .validate_assayDataElementReplace(obj, value): object and replacement value have different dimensions\n An error occured, unable to fit models on data."

This actually occured when I ran KissDE using my data; however, this also resulted after running the example script documented in ?diffExpressedVariants. I also successfully ran qualityControl() and got no error from it.

And I have no idea what and where to look for. Sorry. and thanks for helping out!

Answer 2

[vincent.lacroix]

Dear all,

This bug is now solved. It was related to KissDE's dependency to Biobase and DSS. These packages had been upgraded, and there were compatibiliy issues to fix. The estimation of the dispersion is now also better. Hence, the final output of KissDE 1.4.0 is expected to change compared to KissDE 1.3.0. Some new candidates can be detected. Some borderline candidates are now considered as non significant. The new version can be downloaded here (http://kissplice.prabi.fr/tools/kissDE/). We encourage users to move to version 1.4.0 of KissDE, and 2.14 of DSS. Please tell us if you experience problems.

Vincent