HINGE for minimap

[ebioman]

Hello
I found that intriguing sentence in your paper

"Therefore, integrating HINGE with other overlapping tools such as MHAP or Minimap can be done if different levels of alignment sensitivity or memory usage are required"

Is this a theoretical possibility, something you already started to look into or even possible with current code?

Cheers

[fxia22]

Hi @ebioman ,

HINGE (tentatively) supports minimap, instead of passing --db and --las arguments to hinge filter, maxmial and layout, you can pass --fasta and --paf.

[ebioman]

Hi @fxia22
Thanks for the quick reply, I will give it a (tentative) shot.
Just to be clear, this will mean that there is subsequently no scrubbing possible?

[fxia22]

@ebioman We have our own filtering which contains scrubbing(mainly based on coverage). BTW, the minimap pipeline has not draft assembly and consensus module, so the assembly graph will be final output.

[ebioman]

Hello @fxia22
I was aware of the latter but suspected that some scrubbing was done using DASCRUBBER as you include it in your repository. Thanks for the clarification!

[ebioman]

Hi
I tried a minimal test with minimap and hinge but was not successful yet. I always run into an error with 200 random PacBio RSII sequences.

minimap -Sw5 -L100 -m0 -t20 test.fasta test.fasta > test.paf
hinge filter --fasta test.fasta --paf test.paf -x test --config ../nominal.ini 
terminate called after throwing an instance of 'std::out_of_range'
  what():  basic_string::substr: __pos (which is 18446744073709551612) > this->size() (which is 0)
/software/UHTS/Assembler/HINGE/20170509/bin/hinge: line 8: 38579 Aborted                 Reads_filter "$@"

Update: the strange thing is that if I remove all options and just execute hinge filter I do get the exact same error

If I do the same with DALIGNER it succeeds:

fasta2DB test test.fasta 
DBsplit test
HPC.daligner test | bash -v
DASqv -c100 test test.las
hinge filter --db test --las test.las -x test --config ../nominal.ini 

I used minimap 0.2-r124-dirty

[ebioman]

Sorry to push this, but any ideas what might be wrong - is it working for you in the latest version?

[govinda-kamath]

We have fixed the bug and you can run the pipeline before hinge draft with a paf and a fasta. We'll take some more time to fix draft as paf does not give us detailed alignments for that.

There seems to be a nice work around using minimap2.
This will need lastools and a script for recomputing cigars

The pipeline for the lambda-virus would be the following:

• Download reads

wget https://www.dropbox.com/s/lda4pwsttt168yj/lambda_filtered_subreads.fasta

• Run minimap2

minimap2 -x ava-pb -a lambda_filtered_subreads.fasta lambda_filtered_subreads.fasta > lambda.raw.sam

• Sort and index the sam files

samtools view -bS lambda.raw.sam | samtools sort -o lambda.sorted.bam
samtools index lambda.sorted.bam

• Run recompute-cigar from this repository by Simo Zhang.

python cigar_recomp.py lambda.sorted.bam lambda_filtered_subreads.fasta overlap_recompute <num-processes>

• convert to las using bamtolas:

bamtolas overlaps.las lambda_filtered_subreads.fasta <overlap_recompute.bam 

[ebioman]

Thank you very much, I will give it a try today.

Update:

If I try the original --fasta and --paf combination I still encounter a problem

 hinge filter --fasta subreads.fasta  --paf overlaps.paf -x test --config nominal.ini
terminate called after throwing an instance of 'std::out_of_range'
  what():  basic_string::substr: __pos (which is 18446744073709551612) > this->size() (which is 0)
/software/UHTS/Assembler/HINGE/20170509/bin/hinge: line 8:  3658 Aborted                 Reads_filter "$@"

Is there anything in the config which has to be changed or am I missing something obvious ?

I looked at the work-around but the step of recalculating the cigars is unfortunately very time-consuming and beats the idea of replacing DALIGNER by minimap2 in order to gain speed in assembly.

[govinda-kamath]

Sorry did not see the update to the comment (It looks like we don't get notified about it).

The way we seem to be handling this is to convert the bam to a las. And then run HINGE on the las created (after creating a corresponding db as well). So you'll not have to run hinge with --fasta and --paf

[mictadlo]

Minimap2 has also -c option which stands for output CIGAR in PAF. Is your minimap2 workflow (#112 (comment)) still valid?

If it is still valid after this step bamtolas overlaps.las lambda_filtered_subreads.fasta <overlap_recompute.bam do I have to run

fasta2DB DB -flambda_filtered_subreads.fasta
hinge filter --db DB --las overlaps.las -x asm --config <path-to-nominal.ini>
....

Thank you in advance.