velveth_de running time and mem

[Ori]

Hi,

I have a set of approx 200M reads that I would like to run on velvet.

Could you please share your experience with velveth_de running time and mem consumption?

Thanks

[Ori]

I guess I should be a bit more specific with the question

I've tried to run velveth_de on that set of ~200M reads on local machine (quad core, 12Gb ram). 24h later it is still busy (nothing written to Roadmaps...).

Has anybody similar experience? Tips?

Thanks

[usad]

Hi Ori,

There is the paper of Zhnag et al Practical comparison of de novo genome assembly
also have a look at the wiki how to (http://seqanswers.com/wiki/How-to). This also links to a formula for RAM consumption which was posted in the velvet discussion list
Maybe you can even add in your experience about velvet.

Generally however one would need to know more about your genome coverage etc....

Cheers,
bjrön

[Ori]

Thanks for the info!

I see quite a few velvet runs in front of me, so if I have something interesting, I'll be glad to share

[eslondon]

We have opted for much less memory-hungry algorithms, like Abyss (a bit less friendly), SOAPDenovo, and CLC Bio. Every genome and every set and mixture of read types has different performance depending on the aligner, but at least with CLC Bio, SoapDenovo, etc. you can run many runs, alter parameters, because they are fast and much less memory-hungry...

[Ori]

A little update.

I was running velveth with my set of reads on a bit bigger machine. Still, the memory usage peaked up to 72Gb, filling whole available ram and velvet got very very slow.
So I will try another machine with even more muscles...

I'm planning to try more assemblers during the project.
About velvet and memory its consumption I just curious and intend to figure it out

[ehscholl]

Ori - any updates? I'm running velveth on 12.5M reads, 10M are short, 2M are longer (mostly 454, some sanger).

I did the 10M short reads on my workstation (16GB RAM) with no problem. Adding the extra 2M, I've been running for about a week with all of the RAM taken. I estimate about a week left in the run

[Ori]

Hej ehscholl!

Yes, I have some updates
I filtered my read set rather harsh, reducing total count to approx 80M.
velveth (k-mer 21 to 31) took 18h to complete, memory peaking at approx 90Gb
velvetg took from 4h to 1h to complete, memory peaking at approx 30Gb

Thanks for reminding to share the experience!

[cam.jack]

Ori - it would be very helpful if you said how big you expected your final genome to be and the expected coverage you have (also the length of your reads would be good).

Thanks.

[Ori]

hej,

some info:
expected genome size 18-24mb
expected coverage 10-30x
reads 35bp

[cam.jack]

Thanks for that Ori,