Hisat2 versus Bowtie2

Question

[nalpas.nicolas]

Hi everyone,

I am currently analysing Exome-seq data (paired-end) for two samples. And for the alignment I tried Bowtie2 and Hisat2, with very similar parameters.

The command I have used to run Bowtie2 is as follows: bowtie2 -x $IDX -1 $file -2 $secfile -S ${WKDIR}/${outfile}.sam -q --phred33 --sensitive-local --un-conc-gz ${WKDIR}/${outfile}_failalign_%.fastq.gz --al-conc-gz ${WKDIR}/${outfile}_concoralign_%.fastq.gz --met-file ${WKDIR}/Bowtie2_metrics.txt --no-unal -p $THREADS

and the command for Hisat2: hisat2 -x $IDX -1 $file -2 $secfile -S ${WKDIR}/${outfile}.sam -q --phred33 --un-conc-gz ${WKDIR}/${outfile}_failalign_%.fastq.gz --al-conc-gz ${WKDIR}/${outfile}_concoralign_%.fastq.gz --met-file ${WKDIR}/Hisat2_metrics.txt --no-unal -p ${THREADS} --no-spliced-alignment

I am happy with both alignment, however I notice that Bowtie2 has an alignment rate of ~99% while Hisat2 has ~90% (see detail below):

> /home/nalpas/bin/Bowtie2.sh
Start 14:58:04 20-04-2016.
53505995 reads; of these:
  53505995 (100.00%) were paired; of these:
    9489804 (17.74%) aligned concordantly 0 times
    32491522 (60.73%) aligned concordantly exactly 1 time
    11524669 (21.54%) aligned concordantly >1 times
    ----
    9489804 pairs aligned concordantly 0 times; of these:
      5207409 (54.87%) aligned discordantly 1 time
    ----
    4282395 pairs aligned 0 times concordantly or discordantly; of these:
      8564790 mates make up the pairs; of these:
        918451 (10.72%) aligned 0 times
        2809961 (32.81%) aligned exactly 1 time
        4836378 (56.47%) aligned >1 times
99.14% overall alignment rate

Aligning /media/sf_F_DRIVE/Nicolas/A375_S_R/FiltReads/GS150692_R1_catfilt.fastq.gz and its pair /media/sf_F_DRIVE/Nicolas/A375_S_R/FiltReads/GS150692_R2_catfilt.fastq.gz ...

39440732 reads; of these:
  39440732 (100.00%) were paired; of these:
    5687377 (14.42%) aligned concordantly 0 times
    24753441 (62.76%) aligned concordantly exactly 1 time
    8999914 (22.82%) aligned concordantly >1 times
    ----
    5687377 pairs aligned concordantly 0 times; of these:
      3587160 (63.07%) aligned discordantly 1 time
    ----
    2100217 pairs aligned 0 times concordantly or discordantly; of these:
      4200434 mates make up the pairs; of these:
        348229 (8.29%) aligned 0 times
        1284399 (30.58%) aligned exactly 1 time
        2567806 (61.13%) aligned >1 times
99.56% overall alignment rate
Completed 03:55:07 21-04-2016.


>/home/nalpas/bin/Hisat2.sh
Start 23:40:10 04-05-2016.
53505995 reads; of these:
  53505995 (100.00%) were paired; of these:
    13290978 (24.84%) aligned concordantly 0 times
    34091291 (63.71%) aligned concordantly exactly 1 time
    6123726 (11.44%) aligned concordantly >1 times
    ----
    13290978 pairs aligned concordantly 0 times; of these:
      4551330 (34.24%) aligned discordantly 1 time
    ----
    8739648 pairs aligned 0 times concordantly or discordantly; of these:
      17479296 mates make up the pairs; of these:
        10807086 (61.83%) aligned 0 times
        4454883 (25.49%) aligned exactly 1 time
        2217327 (12.69%) aligned >1 times
89.90% overall alignment rate

Aligning /media/sf_F_DRIVE/Nicolas/A375_S_R/FiltReads/GS150692_R1_catfilt.fastq.gz and its pair /media/sf_F_DRIVE/Nicolas/A375_S_R/FiltReads/GS150692_R2_catfilt.fastq.gz ...

39440732 reads; of these:
  39440732 (100.00%) were paired; of these:
    8671224 (21.99%) aligned concordantly 0 times
    26184600 (66.39%) aligned concordantly exactly 1 time
    4584908 (11.62%) aligned concordantly >1 times
    ----
    8671224 pairs aligned concordantly 0 times; of these:
      2814874 (32.46%) aligned discordantly 1 time
    ----
    5856350 pairs aligned 0 times concordantly or discordantly; of these:
      11712700 mates make up the pairs; of these:
        7800874 (66.60%) aligned 0 times
        2679088 (22.87%) aligned exactly 1 time
        1232738 (10.52%) aligned >1 times
90.11% overall alignment rate
Completed 05:11:54 05-05-2016.

I would have expected that Hisat2 would give me better results, so now I am wondering whether I missed some important parameters in my Hisat2 process. Does anyone have any ideas/experience?

Best wishes,

[Devon Ryan]

I would imagine that the --sensitive-local setting is the main difference. Try trimming your reads a bit more and perhaps the metrics will end up similar. Alternatively, try using local alignment with hisat2 (if you can, I'm not sure that's an option).

Edit: Hisat can do local alignment, hisat2 can't.

[rrcutler]

I got similar results of bowtie2 mapping better (97.42%) than Hisat2 (93.74%) when mapping raw reads (unfiltered) to assembled contigs. Which should I trust more, since it seems Hisat2 can't use a local alignment, as Devon said.