Choosing an Assembler(s)

[charltt]

Hey all,

I am trying to choose the best assembler(s) for a project I am working on. I was looking through the wiki page, and there are so many different choices of assemblers available. I was wondering if anyone would be able to provide any insight into some of the better options.

I will be working on a Linux machine, with a genome size of approximately 64Mb long. The data is in short (40b), single-end reads from the Illumina Solexa Genome Analyzer. I am looking to assemble to a set of reference contigs/scaffolds as well as de novo and compare the results of both. So far, I am looking into using MAQ for the assembly to reference and Velvet for the de novo assembly. Any thoughts?

Thanks!

[Apexy]

Hi,

I think you are on the right track. You may recruit ABySS and RAY subsequently. I think the journey starts with a step so velveth is.

Cheers

[nickloman]

Hey all,
I will be working on a Linux machine, with a genome size of approximately 64Mb long. The data is in short (40b), single-end reads from the Illumina Solexa Genome Analyzer. Any thoughts?
Thanks!

I'd consider giving up now ... your dataset is not really fit for purpose.

[charltt]

I'd consider giving up now ... your dataset is not really fit for purpose.

What makes you say this? There is alot of literature out there about assembling whole mammalian genomes (which are much larger than the one I am working with) from short reads. My biggest concern is that the reads are single-end rather than paired-end. Is this what you are referring to?

[nickloman]

My biggest concern is that the reads are single-end rather than paired-end. Is this what you are referring to?

Yes, I think you really need paired-end reads to even make a start here. Mate-pair reads or longer reads also extremely helpful.

You can validate this easily enough with your dataset; run Velvet with a k-mer of 23 or so and see what the results are like.

[konrad98]

As Nick alluded, a single type of library will never give you as good an assembly as a mix of different insert sizes and read lengths. However single-end reads should be enough to give you most genes depending on repeats, gc bias and the size of introns.

[dawe]

Hi, I have just seen a keynote about SUTTA which seems to outperform most of the assemblers I've used so far.

d