Hi,

I'm doing differential analysis with edgeR. In that package goana function worked well for Gene ontology but kegga is giving some error.

head(de)
           logFC unshrunk.logFC   logCPM       PValue         FDR
58527   2.919535       2.919771 5.010376 4.367288e-06 0.003949316
111    -4.106972      -4.107831 4.357303 1.555287e-06 0.002932518
55969   3.894210       3.895717 3.333907 6.711165e-06 0.004988639
23492  -2.348775      -2.348888 5.475329 4.212049e-06 0.003949316
23268  -2.445401      -2.445518 5.516678 2.048047e-06 0.003181348
143244  3.754845       3.757757 2.245613 4.540069e-06 0.003949316

keg <- kegga(de, species="Hs")

Error in `[[<-.data.frame`(`*tmp*`, s, value = c("0.00394931550141463",  : 
  replacement has 7 rows, data has 30

Any help is greatly appreciated !! thank you.

When you run kegga() in edgeR, the de argument is supposed to be a fitted model object (DGELRT or DGEExact). Please type help("kegga.DGELRT") for documentation.

You have seem to have instead input a data.frame. That will generally give an error because kegga() isn't intended to accept a data.frame. Even it doesn't give an error, the results will be nonsense because kegga will try to interpret each column of the data.frame as a vector of gene IDs.

goana() and kegga() are exactly the same in this respect. If kegga() gives an error, then goana() will also give an error if run on the same arguments.

I don't know what your data.frame represents. It looks like you might have run glmTreat and then extracted the data.frame from the table component of the topTags output. Wouldn't it be simpler to just run kegga and goana in the documented way? You can use

lrt <- glmTreat( ... )
k <- kegga(lrt)
g <- goana(lrt)

Apart from being correct, wouldn't that be easier? It cuts out several steps you seem to have done.