Bowtie2: partial mapping of assembled contigs?

[MartinH]

Hi!

Currently I have a bunch of short read data from different illumina libraries (paired-end, mate pair, single-end) of a bacterial genome. I de novo assembled the read data using different assembly tools and now I want to map the calculated contigs back to a more or less closely related reference genome using bowtie2.

Now I thought that it is possible, to map reads (or in my case contigs) partial to the reference, so for example:

ref:
AAAAAAAAAAGGGGGGGGGGTTTTTTTTTTT
contig:
GGGGGGGGGGXXXXXXXXXX

mapped:
AAAAAAAAAAGGGGGGGGGGTTTTTTTTTTT
-------------GGGGGGGGGG
unmapped:
XXXXXXXXXX

But at the moment, bowtie2 wont map the entire contig and moved it to unmapped.

Did I mis a parameter option or is this scenario not possible with bowtie2?

Thanks!
Martin

[MartinH]

Or an other example:

I have a contig:
XXXXXXXXXXXXXXYYYYYYYYYYYYYY

and my reference:
XXXXXXXXXXXXXXZZZZZZZZZZZZZZZZZZZZZYYYYYYYYYYYYYY

Now bowtie2 would move my contig to unmapped, but maybe there is a insertion in my reference, which is not part of my real genome, so mapping the first half and the second half of the contig would be correct.

[dpryan]

I would suspect that the minimum score is too high for bowtie2 to yield the alignments you want. For the most recent example, how long is the read and how big is the insertion (and what parameters are you using for bowtie2)?

[MartinH]

hm, think thats the problem. I do not really know how big the insertion could be, so it is difficult to determine this in the parameters.

Currently i was using bowtie2 with default settings. My contigs are several thousand bp long, for example the max contig length is ~66Kb.

Maybe bowtie2, developed as a short read mapper, is not the right tool to do such things...

[dpryan]

The issue is just that the defaults aren't really set for mapping large contigs against only closely related genomes. You might actually try bwa-mem, which can deal with the chimeric alignments that likely exist when comparing related bacterial species. I expect that that'll end up working better for you.

[MartinH]

ah yes, I will try bwa-mem, thanks!

EDIT
hm... bwa needs fastq files as query input for mapping, but i have only fasta files holding my contigs... -.-

[mastal]

Try using blast or blat.

[boetsie]

I would suggest to try nucmer (part of MUMmer package). If you have a larger genome, use blasr (part of SMRTAnalysis package from pacbio).

[gringer]

I have used minimus2 to do vaguely similar things like this in the past, trying to map about five different transcriptomes to each other. It probably doesn't work too well for large insertions though. BLAT might work, because it will split matches up into components and allows for multiple insertions per match.