Paired-end RNA-seq: large discrepancy in number of forward versus reverse reads

[rachpetersen]

Hi all,

I am having some issues interpreting the output from the bamtools stats command. I am working with paired end RNAseq data generated from olive baboon vaginal swabs (so we are expecting some bacterial contamination). I mapped the sequences to the olive baboon reference genome using STAR, and then used the bamtools stats command to see how many reads are mapping. The output is a bit perplexing because the samples consistently have a much higher proportion of forward strand reads than reverse strand reads, while the number of R1 reads is equal to the number of R2 reads. I have pasted the output of one file below:

**********************************************
Stats for BAM file(s):
**********************************************

Total reads: 30135821
Mapped reads: 1274337 (4.22865%)
Forward strand: 29498803 (97.8862%)
Reverse strand: 637018 (2.11382%)
Failed QC: 0 (0%)
Duplicates: 0 (0%)
Paired-end reads: 30135821 (100%)
'Proper-pairs': 1272676 (4.22313%)
Both pairs mapped: 1272676 (4.22313%)
Read 1: 15067605
Read 2: 15068216
Singletons: 1661 (0.00551171%)


Does anyone have any ideas of what might be going on here? I've looked at the raw reads and there are approximately the same number of reads in the R1 and the R2 files.

Thank you in advance for your advice!

[luc]

Sorry, I have no idea what could be happening here. Have you checked the quality of both forward and the reverse reads with FASTQC, FASTP or similar?

[rachpetersen]

Thanks for your response, luc!

I tried mapping the R1 and R2 files separately in single end mode, and interestingly, got really high mapping percentages. Using the same sample that I gave the output for in my original posting after running paired end mapping, here is the bamtools stats output for the R1 and R2 files separately:

R1 single end mapping
**********************************************
Stats for BAM file(s):
**********************************************

Total reads: 99555409
Mapped reads: 96735523 (97.1675%)
Forward strand: 51635582 (51.8662%)
Reverse strand: 47919827 (48.1338%)
Failed QC: 0 (0%)
Duplicates: 0 (0%)
Paired-end reads: 0 (0%)

R2 single end mapping
**********************************************
Stats for BAM file(s):
**********************************************

Total reads: 99786078
Mapped reads: 97634672 (97.844%)
Forward strand: 50328230 (50.4361%)
Reverse strand: 49457848 (49.5639%)
Failed QC: 0 (0%)
Duplicates: 0 (0%)
Paired-end reads: 0 (0%)


I'm a little confused by this, because when I check the R1 and R2 files, there are the same number of sequences, yet in this output it looks like there are a different number of "Total Reads" in the two files. I also tried sorting the seqs by name in each file prior to running the paired end mapping again, but that didn't help the issue of low mapping + many more forward strand vs reverse strand reads.

Any advice would be greatly appreciated! Thanks in advance for your help!

[HESmith]

Your reads have been processed in some manner (i.e., filtered by quality) so that R1 and R2 are no longer properly paired. For unprocessed reads, the number of R1 and R2 should be identical, and you should have zero singletons: the stats in your first post indicate otherwise. Sorting by name does not help b/c, as soon as this first singleton is encountered, all subsequent reads are out of register/mispaired.

Apparently, your aligner constrains the R2 search space based on R1 alignment; since R1 and R2 are mispaired, it is unable to find a match for R2 in that space.

The solution is to fix pairing with BBTools Repair, then repeat the alignment. Report back whether or not that solved your problem.

[luc]

I very much agree. Thanks HESmith!

Your reads have been processed in some manner (i.e., filtered by quality) so that R1 and R2 are no longer properly paired. For unprocessed reads, the number of R1 and R2 should be identical, and you should have zero singletons: the stats in your first post indicate otherwise. Sorting by name does not help b/c, as soon as this first singleton is encountered, all subsequent reads are out of register/mispaired. ...