Forward read data look great, paired end and reverse reads look erroneous

[robertocnava]

Hi Felix,
I'm hoping you can help with an issue I'm having with paired end sequencing. I'm sequencing an amplicon that is 350bp long. The first ~300 bases on the forward read match what I expect. Biological and technical replicates match closely. When I run PE alignment, the data follows no pattern I can recognize. PE alignment is low and seemingly random. SE alignment of R2 using --pbat has an alignment rate is fine but the data doesn't make much sense to me. Would you mind taking a look? This is a repetitive region in hg38. I've also attached the unconverted amplicon.

Thank you very much for your help
BISMID-NTG-200-1_S0_L001_R1_001.fastq (2).gz
BISMID-NTG-200-1_S0_L001_R2_001.fastq (2).gz
BISMID-NTG-200-2_S1_L001_R1_001.fastq (2).gz
BISMID-NTG-200-2_S1_L001_R2_001.fastq (2).gz
BIS-MID.txt

BISMID-NTG-200-3_S2_L001_R1_001.fastq (1).gz
BISMID-NTG-200-3_S2_L001_R2_001.fastq (1).gz

[FelixKrueger]

Alright, I took a quick look but didn't go into a lot of detail.

I found that R1 and R2 can be aligned to the amplicon separately, with mapping efficiencies being fairly low (~15%) in the default mode. Efficiency increases to ~30, and 50% when the parameters are relaxed to --score_min L,0,-0.4 or -0.6. Using --local, it goes up to >80%. So there appear to be mismatches to the reference that cause the low mapping in default mode, and I believe this comes from the first 9-10 bp of each read:

Read1

Read2

Indeed, almost 100% of all reads start with these bases; not exactly sure how the experiment was designed, but these residues don't seem to align to your reference.

Indeed, either running trim_galore --clip_r1 10 --clip_r2 10 --paired *fastq.gz followed by a relaxed Bismark run produces a good amount of PE alingments. Also PE alignments in --local mode (>80%) produce alignments that are 10bt shorter on either end, when compared to SE alingments.

I hope this helps?

[FelixKrueger]

Apologies, forgot to describe the last plot:

Shown are Read 1 and Read 2 as single-end alignments, reads on top, and a wiggle-plot below, showing that abrupt start and end of the fragments. The bottom track are local alignments. Of note, start and end are 10bp shorter, as a consequence of soft-clipping the 100% biased positions at the 5' end of R1 and R2.