Output of featureCounts() is not integer

Question

[Francis]

Hi, I using featureCount() to get the reads count for next analysis from sorted SAM file, and I find the read count is not integer. It is not feasible because I want to use DESeq2 to analyse the different genes.

So how to get integer by featureCount() or normalize the output to integer?

thanks.

[Matteo Schiavinato]

As a side comment for you and for future readers on this topic, the fractional counts usually mean that they are expected counts and not raw counts. Simply rounding them up won't be the best choice, because raw and expected counts are different numbers.

The best approach, if you want to use DESeq2 afterwards, is to output raw counts (which will be therefore integers).

[GenoMax]

From the link above.

The help page for featureCounts() says

"fraction:

logical indicating if fractional counts will be produced for multi-mapping reads. If TRUE, a fractional count, 1/n, will be generated for each reported alignment of a multi-mapping read, where n is the total number of alignments reported for that read. countMultiMappingReads must be set to TRUE when fraction is TRUE."

[WouterDeCoster]

So those fractions are indeed expected based on distributing multimapping reads, they're not exact nor raw.

[Francis]

Hi,

I do not run with fraction=TURE & countMultiMappingReads=TURE.

I use

featureCounts -T 8 -g gene_id -p -a Yeast.ercc.gff -o count.txt Yeast.sam

Now the output file is not integer. So if I use fraction and countMultiMappingReads, this should be fraction?

The following comment also said "Simply rounding them up won't be the best choice" and I think so. But I don`t have a good idea to output the integer reads count except rounding tnem up.

thank u for your repply.

[Matteo Schiavinato]

Maybe a workaround: htseq-count outputs integer counts.

[WouterDeCoster]

so does featureCounts with default settings

[GenoMax]

Did you run featureCounts with (fraction=TRUE)? Then you could round the fractions as shown in this thread.