RSEM with HISAT2

[Sbamo]

Hello guys,

I have RNA-sequencing data of around 250 patients with leukaemia.
I have built up a basic pipeline using HISAT2 as my aligner with satisfying results. I want to test differential transcript expression between my samples using RSEM, but I can't get it to work with HISAT2.

I am running the tools on the GALAXY platform and using the instructions provided I used RSEM prepare reference to create the reference files to provide to the aligner. My outputs are the following:

rsem ref name.log
rsem ref name.grp
rsem ref name.ti
rsem ref name.chrlist
rsem ref name.transcripts.fa
rsem ref name.seq
rsem ref name.idx.fa
rsem_ref name.3.ebwt
rsem_ref name.4.ebwt
rsem_ref name.1.ebwt
rsem_ref name.2.ebwt
rsem_ref name.rev.1.ebwt
rsem_ref name.rev.2.ebwt

From what I gather from the RSEM Readme I now have to align my reads using rsem ref name.idx.fa as a reference file. Trying this I get the following error:

(ERR): hisat2-align died with signal 11 (SEGV)
[W::sam_read1] parse error at line 43388
[main_samview] truncated file.

Does anyone have experience using HISAT2 with the RSEM reference file?
It seems to me, that the prepared reference only works with TopHat2 since it creates BowTie index files. I would appreciate any response!

Thanks!
Sbamo

[dpryan]

You can't use those files with HISAT2. You'll need to reindex either name.idx.fa or name.seq (I assume it's a fasta file) and use the resulting .ht2 files.

[Sbamo]

dpryan, thank you for your response! Since I am new to Bioinformatics is there any tool or method you can suggest, in order to convert those files to .ht2-files?

[dpryan]

There are no conversion programs. Just delete them and build the index with hisat2-build.

[Sbamo]

Hi, thanks again for your reply.
It was possible to reindex the "rsem ref name.idx.fa" file (trying to reindex "rsem ref name.seq" threw an error). Unfortunately when using it as a reference HISAT2 presented the following problem:

[E::sam_parse1] CIGAR and query sequence are of different length
[W::sam_read1] parse error at line 42905944
[main_samview] truncated file.
Error while flushing and closing output
terminate called after throwing an instance of 'int'
(ERR): hisat2-align died with signal 6 (ABRT)
[bam_sort_core] merging from 19 files...

I looked up this part: "Error while flushing and closing output
terminate called after throwing an instance of 'int'"
and it seems this is a common problem with the Bowtie/TopHat2 aligners. Sadly, none of the fixes suggested worked and there was no post about this issue in HISAT2.

Please note that the GTF-file and Fasta-file used to create the RSEM reference have been tested and HISAT2 can succesfully run through using them prior to RSEM-conversion (They are Hg19 references).

As always any help is appreciated!

[dpryan]

Can't RSEM just be fed a BAM file? That'd be easier. Anyway, I'd personally just use salmon or kallisto and be done in a fraction of the time

[robp]

RSEM cannot handle alignments with gaps in them; only matches / mis-matches. I would suggest that you use the alignment-based mode of salmon if you'd still like to use the HISAT alignments downstream. Otherwise, you could try using the quasi-mapping-based mode of salmon or sailfish. These programs can produce accurate quantification estimates very quickly without the need to first perform traditional alignment of the reads. Full disclosure: I am the main developer of both of these tools .

[Sbamo]

Robp, thank you for your reply! I may try them out, since for my analysis gapped alignment is mandatory!

[Sbamo]

P.S. dpryan, RSEM can be fed an BAM-file directly, but I had no luck doing this either. Probably due to the fact that it was created using gapped alignment. Thanks for your quick replies!