We are going to prepare a workstation for NGS data analysis. We wanna work with GS FLX+ and Illumina HiSeq 2500 data obtained from gut microbiota of animals. While some part is just 16s rDNA data, the rest would be WGS of gut microbiome.
I was wondering if you could help to make the final decision on one of the below options?
Our options are
1. Intel Xeon E3-1245 v3 Processor (8MB Cache, up to 3,80GHz), 16GB DDR3 RAM, 3,5-Zoll 2TB SATA HDD 7200U/min, 256 GB SATA Solid-State-Laufwerk.
2. Intel Xeon E3-1245 v3 Processor (8MB Cache, up to 3,80GHz), 32GB DDR3 RAM, 3,5-Zoll 1TB SATA HDD 7200U/min, 256 GB SATA Solid-State-Laufwerk.
3. 2 * Intel Xeon E5-2640V2/ 2 GHz (8-Core), 128 GB DDR3 SDRAM, 1TB Standard Serial ATA 600 HDD 7200 rpm.
4. 2 * Intel Xeon E5-2620V2/ 2,1 GHz (6-Core), 128 GB DDR3 SDRAM, 1TB Standard Serial ATA 600 HDD 7200 rpm.
Appreciate your comments!
What will you be doing with your 454 and Illumina data? Whole genome assembly only? Transcript assembly? Mapping reads to reference genomes?
For assembly you typically need massive amounts of RAM, unless your data set and genome are not too large (such as bacterial genomes). You might be able to assemble in some cases with 128GB, otherwise I would agree with Artur and say that computing clusters are the way to go.
For mapping reads with programs like BWA/Bowtie etc. you don't need much RAM (16 or 32GB is fine), and the number of cores available (and their speed) will determine how long your job takes to run.
Also, as Artur mentioned, all hard drive options you mentioned will not be enough to work with large data sets. You will need multiple 2TB or 3TB drives for storage, plus extras for backup.
What will you be doing with your 454 and Illumina data? Whole genome assembly only? Transcript assembly? Mapping reads to reference genomes?
For assembly you typically need massive amounts of RAM, unless your data set and genome are not too large (such as bacterial genomes). You might be able to assemble in some cases with 128GB, otherwise I would agree with Artur and say that computing clusters are the way to go.
For mapping reads with programs like BWA/Bowtie etc. you don't need much RAM (16 or 32GB is fine), and the number of cores available (and their speed) will determine how long your job takes to run.
Also, as Artur mentioned, all hard drive options you mentioned will not be enough to work with large data sets. You will need multiple 2TB or 3TB drives for storage, plus extras for backup.
Since NGS data are high in volume and to assemble them, you will require a lot of RAM as well as Hard Disk space. so from your options, I'll suggest option 3 with 3-4 external hard drives to store your processed data and other back ups.
- 16S/18S analyses of biodiversity can be presently done on any of these computers.
- transcriptomes and metagenomes will require at least 256 Go RAM >20 cpus, and you may easily expect to get 2*250 Go of paired end files. In this case, you want at least 3 To of free space on disk (Trinity for example) for a new analyse.
But if you ask this question, it shows you are not ready for such analyses.
I suggest that you either try to collaborate with some bioinformatician with the proper expertise or that you go and spend at least 1 month in a lab with this expertise.
Finally note that rRNA biodiversity, transcriptomics, genome assembly and metagenomics require very different skills. Metagenomics also requires a bunch of specialists for interpreting the results.
Dear Mahdi,
It boils down to what is volume of data you wish to analysis and also the applications which you choose to analyze. How much data do you wish to generate per run and further per sample. The main reason to ask this is because the choice of assembly tools and allocation RAM and core management is a crucial in this case.
Please feel free to get in touch. [email protected]
Best regards,
Gaurab
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer