featureCounts strand specificity

Question

[tonja.r]

I was working with featureCounts and had three runs while changing only one parameter: the strand specificity. I would expect that the reads from -s 1 and the reads from -s 2 would sum up to the the number of reads outputted by -s 0.

featureCounts --minReadOverlap 50 -s 0 -Q 1 -T 12 -a matrix_gene.saf -F SAF -O -o $OUT/total_counts.txt file.bam
featureCounts --minReadOverlap 50 -s 1 -Q 1 -T 12 -a matrix_gene.saf -F SAF -O -o $OUT/total_counts.txt file.bam
featureCounts --minReadOverlap 50 -s 2 -Q 1 -T 12 -a matrix_gene.saf -F SAF -O -o $OUT/total_counts.txt file.bam

However, the results differ:

==> minReadOverlap50_strand1/total_counts.txt.summary <==
Status    WEN1_6558.sort.bam    WNN1_6545.sort.bam WNN2_6550.sort.bam
Assigned    4643945    8863560    8859072

==> minReadOverlap50_strand2/total_counts.txt.summary <==
Status    WEN1_6558.sort.bam    WNN1_6545.sort.bam  WNN2_6550.sort.bam
Assigned    4775184    9123446    9158397

==> minReadOverlap50/total_counts.txt.summary <==
Status   WEN1_6558.sort.bam    WNN1_6545.sort.bam   WNN2_6550.sort.bam
Assigned    8356549    15881043    15933323

4643945+4775184=9419129 != 8356549

Why don't the reads sum up?

Answer 1

[Carlo Yague]

You misunderstood the role of the -s option:

-s <int>      Indicate if strand-specific read counting should be performed.
                  It has three possible values:  0 (unstranded), 1 (stranded) and
                  2 (reversely stranded). 0 by default.

When you choose either option 1 or 2, reads mapping on both - and + strands are taken into accounts. So it is not expected to sum up. The difference between -s1 and -s2 is that the strands are reversed in -s2. '-' becomes'+' and '+' becomes '-' and as a consequence, '+' reads are assigned to '-' features.

[tonja.r]

I have ChIP-seq data, so I have "+" and "-" reads. I have regions where I want to count my reads. Regions can overlap and can be on "+" and on "-" strands. I do not care about the strands, meaning if any read falls into a feature on "-" strand, it should be counted, if any read falls into a feature on "+" strand, it should be counted. Does it mean that I have to use -s0 and not care about strand specificity?

[Carlo Yague]

Exactly ! -s1 or -s2 are mostly used with strand-specific RNA-seq data. With ChIP-seq data you'll usually use -s0.

[Mithil Gaikwad]

One small confusion is there, I am using paired-end RNA-seq data and for quantification, Platform for my sequencing was 'Illumina NextSeq 500 (Homo sapiens)' I have used the following command

featureCounts -p -t exon -g gene_id -a hg38ucsc.gtf -o SRRXXX.bam.txt SRRXXX.bam

I want to see counts of both the strand (reverse and forward) in my output, Which option I am supposed to add in my command -s 0 or -s 1 or -s 2 option?

[Carlo Yague]

If you want to count on both strand, use -s 0 (which is the default behaviour).

Answer 2

[OD]

Hi,

inspired by the post: A: featurecounts vs samtools view

Maybe your saf file contains overlapping features in different directions. Hence featureCounts will assign reads depending on the -s parameter either to the one or the other feature (i.e. counting the same read twice) but will assign reads only once to one feature in -s 0.

Could you maybe check for those occurrences in your saf file and run featureCounts on only this portion of the file for a test?

[Carlo Yague]

You can also try running featureCounts with the -O option. It will count reads overlapping features once by feature.

[tonja.r]

I did it (in commands above I have -O option) and still something was wring there.

[Carlo Yague]

Yes, my bad, I now know what your issue is. Replying in "answers"

[tonja.r]

As mentioned in the comment above, I can run the featureCount with -O parameter. And I did run it with this parameter (see commands), so I expected that with -s 0 the reads would be counted for each feature, even it they overlap. I also had a test example with two different features on negative and positive strand overlapping one read. With -O and -s 0 the read was counted twice: once for positive and once for negative strand

[Jingjingzhang]

[Jingjingzhang]

hi, with -O and -s 0, the read was counted twice, but if with -O and -s 1, the read will be counted once?