I have run Kallisto on my samples to try and get transcript-level intensities. I now want to do a transcript-level expression analysis using the catchKallisto command from the edgeR package.

This went quite well.

library(edgeR)
paths <- list.files(path = "kallisto/output/", pattern="Sample", full.names=TRUE)
data <- catchKallisto(paths = paths, verbose = TRUE )

But the resulted list$counts are not counts (Are these TPMs???).

             kallisto/output//L75442_Track-118107
Y110A7A.10.1                         1.439992e+03
Y110A7A.10.2                         8.353787e-03
F27C8.1.1                            1.200000e+01
F27C8.1.2                            0.000000e+00
F07C3.7                              5.300000e+01
             kallisto/output//L75443_Track-118108
Y110A7A.10.1                          1575.499243
Y110A7A.10.2                             1.500757
F27C8.1.1                                0.000000
F27C8.1.2                               11.000000
F07C3.7                                 52.000000
             kallisto/output//L75444_Track-118109
Y110A7A.10.1                         1.546000e+03
Y110A7A.10.2                         1.515167e-04
F27C8.1.1                            5.000000e+00
F27C8.1.2                            0.000000e+00
F07C3.7                              7.000000e+01

Do I need to change them into counts (just by rounding them) or is it ok for edgeR to have them as such?

thanks

Assa

They are estimated counts from kallisto's EM algorithm. Columns sum to the total number of mapped reads for that sample.

edgeR accepts fractional counts, using a continuous extension of the negative binomial distribution: https://stats.stackexchange.com/questions/310676/continuous-generalization-of-the-negative-binomial-distribution/