Dear all, I want to run the kissDE R package on my set of alternative splicing events that I outputted from KisSplice (…_coherents_type_1.fa='results_allgreg.fa').
I'm using this code:
myCounts <- kissplice2counts('results_allgreg.fa',pairedEnd=TRUE) # load kissplice data mySplicingconditions <- c("t1","t1","t1","t4","t4","t4","t7","t7","t7","t10","t10","t10","t13","t13","t13","t16","t16","t16","t19","t19","t19","t22","t22","t22")
# add 1 to all counts - cannot have any 0s and all reads need to be integers
class(myCounts)
head(myCounts[[1]])
myCounts_new = myCounts
myCounts_new[[1]][1:54348,3:26] = myCounts_new[[1]][1:54348,3:26] + 1
head(myCounts_new[[1]])
# remove NAs and corresponding rows in element 2
myCounts_noNA = myCounts_new
myCounts_noNA[[1]] = na.omit(myCounts_new[[1]]) #remove NAs from element 1
head(myCounts_noNA[[1]])
myCounts_noNA[[2]] = head(myCounts_new[[2]],-5)
diffSplicing <- diffExpressedVariants(myCounts_noNA, mySplicingconditions)
write.table(diffSplicing$finalTable, file="gregkissDE_output_finalTable.txt", sep="\t", quote=FALSE) # the finalTable output file is empty
kissplice2counts created a large list of 4 elements that included several zeros. I notice that in the updates list Version 1.0.0 kissDE version 1.0.0 (2014–10–29) includes “a method to bypass model calculation issues due to counts equal to zero in data”. However, I cannot find the option to handle zeros.
I have attempted to overcome this issue by changing all zeros to 1 and removing NAs as well, but I get the error below when running diffExpressedVariants. diffExpressedVariants creates a large list with 2 elements but the output diffSplicing$finalTable is empty.
"Error in data.frame(..., check.names = FALSE): arguments imply differing number of rows: 13387, 10711\n Returning only resultFitNBglmModel and sing.events"
How do I use diffExpressedVariants with some zero counts in my data?
Any suggestions will be appreciated. Thanks