featureCounts parameters for RNA-Seq

Question

[CuriusScientist]

Dear Community,

I have been running featureCounts to count mapped reads in RNA-Seq paired-end data with the following parameters

featureCounts -T 10  -t exon -s 2 -p -g gene_id -a <path_to_annotation.gtf> -o <path_to_outoput_directory> mapped.bam

as mentioned in http://bioinf.wehi.edu.au/featureCounts/

However, I came across two posts where the parameters

• --primary https://bioinformatics-core-shared-training.github.io/CRUK_CI_RNAseq_course/Supplementary_Materials/S2_Read_Counts_with_Subread.html

and

• --largestOverlap How to calculate distribution of mapped read over classes of RNA?

have been used.

This got me a little confused and will like to to know if and when I should use these parameters as well as what will be the difference in the output from the default parameters?

Many thanks in advance

[wang-yanfang]

Hey do you get some answers for this? I am also using featurecounts now, paired ends RNAseq, and I am not sure if i have to use specify strandness? Could you please help me ?

[Gordon Smyth]

If your sequencing is stranded then you need to specify to featureCounts which strand has been sequenced. If your sequencing is not stranded then you don't. Pretty logical!

[wang-yanfang]

Thanks for your answer. May I ask do you reccomand --primary to be set on?

[Gordon Smyth]

Generally, we recommend that featureCounts be run with default settings, which does not include --primary. Just follow the documentation. The specific use depends on your intended application.

[wang-yanfang]

Thanks so much Gordon Smyth. In default setting -t exon is set on. So when is -t gene reccomanded? In my bacteria gff file, I do not see exon, but in my fungi gff file, I do. I wonder can I just set -t gene for both bacteria and fungi? I really appreciate your help. Best regards, Yanfang

Answer 1

[swbarnes2]

Honestly, if you aren't using something like Kallisto, which outputs counts more or less natively, I'd be using RSEM, not featureCounts RSEM is smarter about dealing with ambiguously assigned reads.