Demultiplexing a MiSeq library based on Amplicon Sequence divergence?

[SDK]

Hi All,

My lab works on a lot of diverse taxa. We want to sequence the mtDNA genomes of each using our MiSeq. We generated long-range PCR amplicons for each species, then pooled 10 taxonomically divergent species into each library (total Miseq run was 10 libraries containing 10 species in each library). Each library is labelled with a Nextera ID. We planned on demultiplexing each library based on the non-similarity in sequences. The run was successful, and now it's time to demultiplex the libraries. We had a guy that was going to do this for us, but he's super busy so we are going to try it ourselves...but we don't know how we should go about this. Any suggestions regarding programs or how to go about it? Since each library contains only amplicon sequence can we simple ask MIRA to build contigs? Any help is appreciated.

Cheers,

[AdrianP]

All you want is mtDNA right?

Just assemble it w/o demultiplexing, you should get mtDNA contigs for each of your taxa.

How long is your fragment size? With MiSeq + your situation I recommend MIRA.

[SDK]

Thanks Adrian. The mtDNA genomes are between 15-18 kb, however we had multiple PCR's of a smaller size (~5-10 kb) to generate a full genome, then we pooled it together. Basically, my pooled library contains mtDNA amplicons of 10 distantly related species (i.e., a bat, a fish, a bird, etc). The 2 closest species are both fish, but still very distantly related. So you think MIRA can just assemble them denovo by asking for contigs to be built? No need for reference genomes (some of which we have, some not)? Thanks for answering!

[AdrianP]

As long as your PCR product are overlapping, you should be fine. Indeed you can just denovo assemble it by feeding it the entire library. Assemble each library individually.

As a side thing, you might want to check the quality of your data. You said you have some reference genomes. Use bwa to map a library containing the mtDNA of the reference genome to the reference genome sequence, then visualize read coverage with tablet, it will tell you if the entire mtDNA is covered. It's like quality control. If some regions are poorly covered or not at all, then shucks, you might not get your circular contigs.

[SDK]

Thanks Adrian, I'll give that a try. Have you ever tried this before? Or heard of it begin done? Would love to hear some 'proof of concept' testimonial!
Cheers,

[AdrianP]

I have never heard of your initial idea of demultiplexing based on similarity. That's essentially what an assembler does, separates reads based on which reads form contigs...

There is no proof of concept here, it's plain and simple, you have metagenomics, multiple genomes in one library. Assemblers handle that well. It really depends on the depth that each mtDNA got sequenced at.

[mastal]

The assemblers should be OK when the species are not closely related, if the species are too closely related you may get chimeric contigs.

You may have problems due to uneven coverage, even if you try to pool equimolar amounts of each amplicon, in practice it never comes out quite equimolar, and some amplicons will be present at higher concentrations than others. This causes problems for some assemblers which assume even coverage.

Do adapter trimming/quality trimming of the reads before assembly.

As mentioned above, MIRA is a good assembler to try.