Create_Local_Contig_Kmer_Index(GraphConstruction.h) SegFault about dbg2olc HOT 7 CLOSED

Result 5 means the two sequences do not align, however, they were recognized to align in an earlier stage, as you have this bug in the final backbone construction step.
I wonder if the program is called in an advanced way?

from dbg2olc.

Hey @yechengxi,

I have ~38x of RSII data with read L50 of 18.8kb, reads 3kb+ only. I ran DBG2OLC with the following parameters:

DBG2OLC LD 0 k 17 KmerCovTh 2 MinOverlap 20 AdaptiveTh 0.005 RemoveChimera 1 MinLen 3000 Contigs contigs.fa f subreads.3kb.fasta

This produced a segfault. Varying the AdaptiveTh parameter also produced segfaults on some (but not all) runs.

from dbg2olc.

Hi Bredeson,
I see no problem in your command, but may I know which NGS assembler did you use for the Illumina assembly?

from dbg2olc.

Hey @yechengxi,

I used contigs generated by 10X Genomics' Supernova, as they were more complete than contigs I could generate with other assemblers. The organism I'm attempting to assemble is highly inbred, by the way.

from dbg2olc.

This could be the cause of the seg fault. I have no idea if other assemblers had heuristics that create unexpected structures for DBG2OLC. But you may be able to dig further into the problem as you did in the head of the thread. I would say you are pretty close to a solution.

from dbg2olc.

@yechengxi,
Would it be sufficient to include a conditional statement to check if the result is 5 and return from Create_Local_Contig_Kmer_Index() to handle this case? On that note, however, I do see in my log file that other alignments have returned result=5 without segfaulting...

from dbg2olc.

If I remember correctly (about this 2 yr old project) if the result is 5 we still need to use the alignment to build the backbone.
Result 5 means the alignment is poor. But still, we need to look at the alignment result to join the reads. This means, find a shared contig, find some shared k-mers and join somewhere.
I just wonder what is happening in your environment.

from dbg2olc.