No Bowtie Alignments

[sewellh]

A former lab member assembled a number of contigs from Illumina reads using SPAdes. I have been trying to assess the depth of coverage using Bowtie2 when I noticed something interesting. I find that there are no Bowtie alignments (concordant or discordant) for the largest contig. Can anyone explain this?

[GenoMax]

Did you try taking (parts of) the large contig and blasting at NCBI to confirm if the data is from the same/similar species as you expect it to be? Are you mapping the same data back to the contigs?

[sewellh]

I am mapping the reads used to make the contigs back on to the contigs. I get alignments for all contigs except for the largest one. I have BLASTED the contig and it is what I expect.

[GenoMax]

I am mapping the reads used to make the contigs back on to the contigs. I get alignments for all contigs except for the largest one. I have BLASTED the contig and it is what I expect.

Not sure why you are not getting any hits to the largest contig. Have you redone the reference indexing to verify that the contig is included (and there were no errors with the index creation) and/or tried a different aligner?

[sewellh]

Yes, I redid it. Someone mentioned that SPAdes does error correcting on the reads prior to assembly which might result in differences, so I'm trying today to run the error correction on the raw reads before the Bowtie2 alignment.

[GenoMax]

It still seems odd that it is affecting only that contig and nothing is aligning. How are you determining that nothing is aligning (by inspecting the BAM)?

[sewellh]

Yeah, by inspecting the SAM file.

[Brian Bushnell]

1) Have you verified that all of the contigs have unique, correctly-formatted names?
2) Does the contig look normal to you - high complexity, mainly defined bases, rather than e.g. a homopolymer or mostly-N sequence?
3) Is it possible that this contig is a replicate of other contigs? Even though it's bigger, it could be fully covered by other contigs. So, do any other contigs map to it?
4) Is it highly repetitive such that reads aligning to it might exceed the maximum number of allowed alignments?

[sewellh]

Yes, the contigs have uniqe and correctly formatted names. But even when I try to just map the reads to the single large contig, I get no matches.

It doesn't look like this contig is a replicate of others but it does have a 3-4 copies of a ~500 nt fragment within itself. Does that mean that this contig was made incorrectly or that there is something else I should do? I would would expect that if I tried to align the raw reads just to single contig that I would get some alignments.

Update: Using the resulting fastq files from the Hammer error correcting, I still get no Bowtie alignments to that contig

[Brian Bushnell]

Sounds like a bug, but I don't know if it's in the assembler or aligner. I suggest trying a different aligner such as BBMap; and be sure to set the "ambig=all" flag so that reads without a unique best mapping site get mapped to all top-scoring sites.

[sewellh]

Thanks so much for your help. I'll try out BBMap. If you are curious at all to look at the contigs, they're on JGI. The largest is:
>gi|589096183|gb|JARN01000011.1| Dehalococcoidia bacterium DscP2 WGS:JARN01:comHGAPfinal_Contig11_1.11, whole genome shotgun sequence