A server for processing data from Nextseq 500？

[johnsunx1]

Hi everyone,

We just buy a nextseq 500 and plan to buy a new server for processing the sequencing data, including exome-capture, RNA-Seq, ChIP-Seq, etc. Our budget for server is around $100K, and we are hunting for the server brand, configuration, etc. Do you have experience with nextseq 500? Any suggestion for the server is highly appreciated!

Thank you!

John

[GenoMax]

If you have no resident IT expertise then my first suggestion is to get your local/central IT to manage this server for you. Buying a server is the easy part, ongoing maintenance/administration can be a chore, especially if you are not familiar with this aspect. If you are going that route you could consult them about what vendors they work with.

You have ample budget so no worries there. You could look at buying a single beefy server (with a TB of RAM) or a couple of servers (with one for pre-processing data and one for doing analysis).

Are you only going to pre-process data or expect to do some additional analysis as well? Are you a core who expects to run the machine 24x7 or something not as frequent?

[gringer]

plan to buy a new server for processing the sequencing data, including exome-capture, RNA-Seq, ChIP-Seq, etc. Our budget for server is around $100K

For a couple of runs every month, I can do all of that on a $4,000 server (or desktop, if you want to call it that) with money to spare.

Unless you're a large core facility (and I can't imagine that would be the case if you're only going for a NextSeq), it's unlikely that you'll need anything close to a $100k budget for a server. Spend that money to hire a good bioinformagician who can talk to you about what you actually need for your experiments.

[SNPsaurus]

I mostly use a $3700 dollar workstation for aligning NextSeq runs to a reference, sometimes assembly of small sub-genomes, etc. The only time I ran into a problem is using freebayes genotyper on a polyploid... ran out of memory.

The local facility has a server that was maybe $30k. No one wants to use it after 4 years because the individual CPUs are much, much slower, and any step not optimized for parallel execution is very slow. The university has a million dollar cluster set up as a cloud compute environment. Same issue.

$100k can buy lots of time on Amazon. I'd think of a hybrid strategy and get something in the $10-20k range locally and save money for high-memory nodes on Amazon if needed. I know most funding sources won't allow it, but if you can park any of the money to replace the $10k machine, or for salary support for maintenance and IT support (install packages, for instance), you'll have a much more productive experience than spending $100k on just hardware.

[johnsunx1]

Thank you all for sharing your experience with nextseq!

GenoMax, we plan to run all the analysis in this server. For example, for RNA-Seq, we will run tophat + cufflinks/edgeR; for ChIP-Seq, we will trun MACS, etc. At this time, we assume we will run 24*7. So the server should have enough space and memory to deal with the sequencing data.

[GenoMax]

Since you have the budget available it may be good to go with two (or three) servers to provide redundancy. As pointed out by @SNPsarus and @gringer you can do the pre-processing (base-calling/demultiplexing) of the data on a dual socket workstation/smaller server (8 cores should be enough, I would recommend getting 64G of RAM) and then purchase a big server for downstream analysis. If you can tap into high performance storage from someplace (e.g. isilon, NetApp) it would be an ideal solution. Otherwise getting a direct- or network-attached storage array would be second best option.

If this is a one time money then don't skimp on storage/memory and a good data backup solution (easy to forget).