Counting paired-end sequencing mapped reads

Question

[heir_of_isildur88]

Hi all,

I have a very basic question here. With a paired-end sequencing, how do we count the number of mapped reads?

I did a flagstat on my file which I have already filtered with the flags 83 & 163 for mapped proper and properly paired.

The flagstat results are as below:

69640 + 0 in total (QC-passed reads + QC-failed reads)

0 + 0 duplicates

69640 + 0 mapped (100.00%:-nan%)

69640 + 0 paired in sequencing

34820 + 0 read1

34820 + 0 read2

69640 + 0 properly paired (100.00%:-nan%)

69640 + 0 with itself and mate mapped

0 + 0 singletons (0.00%:-nan%)

0 + 0 with mate mapped to a different chr

0 + 0 with mate mapped to a different chr (mapQ>=5)

So from the results, does 69640 means that 69640 reads mapped, or 69640 paired-reads mapped (as in R2---R1 is counted as 1 read)? Should we divide the number with 2 if only one end of the reads were counted?

A bit confused here. Hope someone can help. Thank you very much.

Answer 1

[Pierre Lindenbaum]

• 69640 reads where mapped
• 69640 reads where mapped and they were part or a paired-end experiment/assay
• 34820 came from the R1/forward fastq
• 34820 came from the R2/reverse fastq
• 69640 read are correctly mapped with their mate (paired-end experiment, good distance, same contig...)

[heir_of_isildur88]

Thanks Pierre for the reply.

But I still don't quite understand. I attached a picture to make my query clearer.

image url

In the picture, the flagstat showed 6 reads; 3 from R1 & 3 from R2. I understand that. But if you view the sam file in a genomic viewer like IGV, the first image appears. But if I view them as pairs, it will look like the second one. So in the second picture, we actually only have 3 fragments of the gene mapped, with each fragment made from a pair of reads. So shouldn't we divide the number of reads by 2 to get the total number of fragments mapped? The number given in flagstat gives the number of individual reads, without taking into consideration the pair, right?

[Pierre Lindenbaum]

So shouldn't we divide the number of reads by 2 to get the total number of fragments mapped?

yes, and this number would be "properly paired" /2

The number given in flagstat gives the number of individual reads, without taking into consideration the pair, right?

again, the number of 'correct fragments' would be "properly paired" /2

[heir_of_isildur88]

OK, that clarifies a lot of things.

But then, what about when the "properly paired" number is odd? It's not always an even number, right? For example, R1 can map to 2 different R2 fragments, how do you count those?

I am quite new to sequencing, so I would like to know what is the general consensus on reporting the number of reads in paired-end sequencing? The total number of reads or the one divided by 2 i.e. 'correct fragments'?