bacterial genome assembly on Miseq

[Etherella]

Hello!
I am planning to make a Nextera XT library prep (a regular bacterial genome, a couple of megabases), sequence it with 300 cycles kit(2x150 bp), and then assemble. As far as I understand miseq has inbuilt software that does everything (including velvet assembly for small genomes) automatically. do I need to launch the process after the sequencing or is everything done automatically?

[TonyBrooks]

Everything should be done automatically if you set it up on your sample sheet. The assembly is carried out in BaseSpace. Although, if you go into the Run Options screen on your MiSeq, you have the ability to replicate the analysis locally - which you might want to do if not using BaseSpace.
We've found the Velvet assembly on MiSeq to be rather hit and miss in terms of quality, ranging from acceptable to very poor. We got much better assemblies using an OLC assembler than we ever got with Velvet.

[mcnelson.phd]

I'll echo Tony's statements about assemblies coming straight off the MiSeq as being very hit or miss. For one genome we once got a 2.8Mbp contig that was nearly perfect out of a 3.5Mbp genome, but we've also gotten assemblies with N50s of 2Kbp and no contigs larger than 50Kbp. A large part of the problem is that the data doesn't appear to be pre-processed in any way to trim off low quality regions or look for PCR duplicates.

I'd suggest setting up your run to produce the assembly, but also do the work yourself to compare. Most likely you'll find that you can do a much better job and be glad you didn't just rely on the system to give you an assembly.

[TonyBrooks]

We once got an N50s less than the read-length (251PE).

<AssemblyStatistics>
<NumberOfContigs>59444</NumberOfContigs>
<MeanContigLength>56.10188</MeanContigLength>
<MedianContigLength>46</MedianContigLength>
<MinContigLength>31</MinContigLength>
<MaxContigLength>560</MaxContigLength>
<BaseCount>3334920</BaseCount>
<N50>62</N50>
</AssemblyStatistics>

No idea what was going on there. All quality stats suggested it was good sequencing (12m reads from v2 500 cycle with 93% >Q30). We assembled the data offline without problems. N50's went up to 150kb.

If you want to use velvet, Nick Loman has a good guide about how to pre-process

http://pathogenomics.bham.ac.uk/blog/2009/09/tips-for-de-novo-bacterial-genome-assembly/

[Etherella]

Thanks a lot! Well, anyaway Miseq stores unaligned fastaq data, so I will be able to have a look at the automatic assembly and then, if the quality is lacking try other software or run Velvet again but with pre-process.

[nucleus]

Ive had trouble with Velvet before. The issue was not the read quality but rather the sequencing that was too deep (>~50x Velvet falls apart). I am now almost exclusively using Spades (http://bioinf.spbau.ru/spades/) which does the read corrections and assembly on one go, and dosent mind very deep coverage. Spades also gives me better results than CLC.