Hi,

1) RSEM input

Having counts of RSEM for several samples, we compiled a table like described in e.g. the Trinity description into one table for all samples. Striuggeling now in DeSeq2 with the function "deseqtable <- DESeqDataSetFromMatrix" we tried now the tximport package reading the manual but again struggeling how to adjust the example "files <- file.path(dir, "rsem", samples$run, paste0(samples$run, ".genes.results"))"...  which takes the counts from the RSEM output tables directly and somehow accounts for non-integers.

Isn't there an alternative way to e.g. import a table from all samples directly (RSEM_trans_iso_counts.counts.matrix)

in the following format

    S1   S2    S3    S4
TRINITY_DN38912_c0_g1_i1    0.00    0.00    1.00    2.00
TRINITY_DN24712_c0_g1_i7    54.45    0.00    17.39    0.00

without rounding it first ?

Would be helpful if anyone can post some helpful command lines :)

2) is DeSeq2 by default excluding results with low counts (see 1.5.3) - how can I set different treshholds or turn off this filtering, by setting p-values set to NA ? (I got that a more strict filtering is
automatically applied via independent filtering on the mean of normalized counts
within the results function (section 3.8) how to change this?

thanks for help

Karen

Can you explain what went wrong when you followed the RSEM example in the tximport vignette? Exactly what errors, etc.

DESeq2 only keeps the integer part of the values. There is no reason to be concerned about loss of precision in rounding, given that these fractions are dwarfed by the shot noise from sampling molecules.

"how can I set different treshholds or turn off this filtering" 

See the DESeq2 vignette or help pages, e.g. ?results. This is discussed there.