computer for RNA-seq analysis

[eggplant72]

Hi,
My lab is looking to get a new computer and I am trying to find something suitable for RNA-seq analysis. We have a few projects in the lab in which we hope to analyze transcriptomes with a reference genome, with a few lanes of Illumina Hi-SEQ data. We are looking for differential expression.

Does anyone have a suggestion of what kind of specs the computer would need (processors, RAM, etc) ?

Thank you

[mhampton]

Hi,
My lab is looking to get a new computer and I am trying to find something suitable for RNA-seq analysis. We have a few projects in the lab in which we hope to analyze transcriptomes with a reference genome, with a few lanes of Illumina Hi-SEQ data. We are looking for differential expression.

Does anyone have a suggestion of what kind of specs the computer would need (processors, RAM, etc) ?

Thank you

Hi,

I do quite a bit of similar work, I am currently shopping for a new computer.

With a reference genome, you don't need lots of RAM usually, but in some situations you might want to do some denovo sub-assemblies (e.g. if you are working with mutants or a different strain). Then you might need more. More RAM also lets you exploit multiple cores in some parallelizable situations.

The RAM choice is important because of choosing between a "consumer" workstation with non-ECC RAM using a Sandy Bridge chip like the Intel 2600K, or going with a server that can use ECC RAM with Opterons or Xeons. Usually the consumer boards max out at 32 GB, but that would probably be enough for what you are doing.

It really depends on your budget. If you can manage more than $4000 or so, I would recommend getting a server with ECC RAM. Something like this perhaps: http://www.avadirect.com/workstation...asp?PRID=13597
beefed up with an SSD drive and two processors.

If anyone has some tips about good places to order such a machine I'd appreciate it (I have not ordered from avadirect before).

[gringer]

I'm managing with 24GB ram, and 12 Xeon X5650 processing units. The de-novo things I've trying to do are too memory hungry for the computer (except for Ray, which is impressively memory efficient), but mapping to reference genomes is fast enough for me (overnight for ~500M sequences).

Trinity is having trouble with the data in terms of memory usage, but I've still got a few things up my sleeve to simplify the sequences and reduce Trinity's workload.