How to align contigs without reference genome?

[yangfangisok]

I have 53 samples which belongs to two treatments. In each sample, I started with 2X100 Illumina metagenomic data and have used SOAPdenovo2 to generated contigs. I want to combine all contigs (>10000bp) from all samples and align all contigs. I want to have alignment map or contigs map. Final goal is to find contigs which are unique to each treatment. I have no idea how to do it? Could anybody help me?

[Brian Bushnell]

One of my programs, Dedupe (packaged together with BBMap), is designed to find duplicate and unique contigs. You can run it like this:

dedupe.sh in=assembly1.fa,assembly2.fa,assembly3.fa out=unique.fa uniqueonly=t

Optionally, you can allow some mismatches, edits, or set various other constraints.

That will output all of the unique stuff and throw away anything with any duplicates. You should first rename the contigs so you know which assembly they came from. If you only have 2 treatments, I would suggest first merging all of the reads from a treatment and assembling them together so you only have 2 assemblies (if I correctly understand your problem).

[flxlex]

You could also try a 'superassembly' using cortex: http://cortexassembler.sourceforge.net/

[yangfangisok]

Thanks. Cortex assembles multiple eukaryotic genomes. But I work on microbiome.

[yangfangisok]

Thanks, Brian.
I'm trying to align contains from different metagenomes to define a “Core” and looking to do at the contig level for richer genome organization and origin context.
Do you know how to do it?

[Brian Bushnell]

Hmmm... other than dedupe, I don't really have any relevant tools. I'm not sure how to best go about that; you might need to write something yourself.

As for dedupe, you could process all the assemblies together, allow some edit distance, set "uniqueonly=t", and set the "outd" flag to save all of the duplicate contigs. For example:

dedupe.sh in=assembly1.fa,assembly2.fa,assembly3.fa outd=duplicate.fa uniqueonly=t e=10

The run dedupe again on the duplicates:

dedupe.sh in=duplicate.fa out=core.fa e=10

This is not a perfect solution, but it will give you a file with exactly one copy each of all the contigs that appear at least twice in the input.

-Brian