HISAT2: Unaligned Reads do not show in paired-end mode only

[FelixKrueger]

Description of bug:

We found that MultiQC seems to miss out the Unaligned reads statistics for HISAT2 in paired-end mode (PE) only, but works fine in single-end mode (SE). Bowtie2 statistics are gathered correctly in both cases.

Here is an example SE plot:

And here is the PE version:

This can obviously be quite misleading at times.

This happens despite the logs looking fairly similar (the programs were run with somewhat different

Bowtie2 report

2022172 reads; of these:
  2022172 (100.00%) were paired; of these:
    494895 (24.47%) aligned concordantly 0 times
    925333 (45.76%) aligned concordantly exactly 1 time
    601944 (29.77%) aligned concordantly >1 times
    ----
    494895 pairs aligned concordantly 0 times; of these:
      105774 (21.37%) aligned discordantly 1 time
    ----
    389121 pairs aligned 0 times concordantly or discordantly; of these:
      778242 mates make up the pairs; of these:
        444082 (57.06%) aligned 0 times
        204517 (26.28%) aligned exactly 1 time
        129643 (16.66%) aligned >1 times
89.02% overall alignment rate

HISAT2 report

2025528 reads; of these:
  2025528 (100.00%) were paired; of these:
    464484 (22.93%) aligned concordantly 0 times
    1458868 (72.02%) aligned concordantly exactly 1 time
    102176 (5.04%) aligned concordantly >1 times
77.07% overall alignment rate

File that triggers the error:
PlateE_H12_S192_merged_L001_GRCh38_bowtie2_stats.txt
PlateE_H12_S192_merged_L001_GRCh38_hisat2_stats.txt

MultiQC run details (please complete the following):

• Command used to run MultiQC: multiqc .
• MultiQC Version: 1.9
• Operating System: Linux
• Python Version: 3.7.3
• Method of MultiQC installation: manual

Additional context
Thanks very much!

[ewels]

I’m confused - the screen grabs both look fine here again? The category names are different for SE and PE, but both have unaligned reads (SE not aligned, PE neither mate aligned).

What are you expecting to see here?

[ewels]

Ahh sorry, I see - it’s the HISAT2 logs mixed in with bowtie2. Gotcha.

[ewels]

Ok, fixed I think. Very hacky though, as is most of the Bowtie2 parsing code. Please test and test and test to make sure that I haven't messed anything up - I find it extremely difficult to tell whether the numbers are adding up properly.

Reminder that you can run HiSAT2 with the --new-summary option. This gives a more sane YAML output file.

[FelixKrueger]

Thanks Phil, I have now added --new-summary to our pipelines, will do some testing soon.