help needed: reference guided assembly of bacterial genomes?

[jerrybug109]

Hi all! Bioinfo newbie here!

We recently used a core facility to sequence the genomes of 40 strains of one species of bacteria (150bp short reads, paired end, illumina) and now I'm trying to figure out the best approach to assemble their genomes.

There are reference genomes out there for this species so I assume that a reference guided assembly would be the best way to go.

I'm trying to decide which genome assembly software would be best to assemble bacterial genomes (with a reference). It seems like several of the "big" ones out there like spADES or Velvet are good at de novo assembly.

What's the difference between genome assembly and alignment/mapping? Why can't I use an alignment like Bowtie to do reference guided assembly? It seems to me that reference-guided assembly and alignment would do the same thing?

Would love your insight, thanks.

[GenoMax]

SPAdes (or some would recommend velvet). If you have enough coverage (and the data is good quality) start there. If you don't get good results then you could start looking at other options.

If you have a good reference you could just align and see what that gets you.

I also would recommend that you look at Mauve for genome wide comparisons.

[jerrybug109]

Thanks!

Someone has also suggested to me that I could de novo assemble all 40 strains and then map those contigs onto a reference genome afterwards?

Is this sort of "hybrid" approach preferable over just trying to do an alignment? It seems like there's numerous approaches out there but I'm not well versed enough to gauge general consensus yet.

Our later downstream goal with the assembled genomes is to look for genetic variation between the strains.

Thanks!

[GenoMax]

Are your genomes true strains of the reference (i.e. you expect them to align with high % identity across the entire genome)?

In your case you can probably tackle the ultimate goal of SNP calls from more than one end but if the answer to the question above is yes the doing alignments to reference followed by SNP calls may be the most straightforward way. If you start seeing gaps/unaligned reads then that could give you an idea of how much the strains vary from the reference. At that point individual assemblies can be tried for a more complete picture. Followed by Mauve analysis.

[jerrybug109]

Are your genomes true strains of the reference (i.e. you expect them to align with high % identity across the entire genome)?

In your case you can probably tackle the ultimate goal of SNP calls from more than one end but if the answer to the question above is yes the doing alignments to reference followed by SNP calls may be the most straightforward way. If you start seeing gaps/unaligned reads then that could give you an idea of how much the strains vary from the reference. At that point individual assemblies can be tried for a more complete picture. Followed by Mauve analysis.

Our genomes are individual strains of bacillus subtilis and the reference genome is bacillus subtilis, so I think my answer to your question is yes.

Thanks.

Since the genomes are only ~4.2 megabases long it seems like it would take a fairly trivial amount of running time to assemble them using spADES or velvet/velvetoptimiser. Maybe we'll go ahead and just do the individual assemblies anyway. I'll try it out a couple ways!