I agree with Dr. Guarracino and will be more focused in the hardware specifications rather than the brand itself. If you are going to do RNAseq analysis the first question would be: Do you have a reference (for example Arabidopsis or rice) or do you need to do a de-novo assembly ? How much data do you have ? What tools do you want to use ? Do you have any previous knowledge of Linux and command line tools ? Here some orientation:

+ If you have a reference, you will not need a big server/workstation. Generally you can follow the protocol detailed in Trapnell C. et al. 2012 (http://www.nature.com/nprot/journal/v7/n3/full/nprot.2012.016.html). Bowtie2 needs ~2.2Gb to map reads to human genome size. More cores will speed up the mapping, but they are not strictly necessary. Tophat2 and Cufflinks are usually less demanding than Bowtie2 so I think that desktop with 8 or 16Gb should be enough in most of the cases. The downstream analysis using R could be a problem if you have a more than 3 conditions to compare. An alternative may be to use Galaxy and load the data to the Galaxy servers for the analysis (http://galaxyproject.org/) but again it will depend of how much data do you have.

+ If you don't have a reference, you will need to create one doing a de-novo transcriptome assembly using Trinity, ABySS-trans, SOAPdenovo-trans or any other assembler of your choice. In any case you will need a big server with at least 64 Gb of RAM memory (128 Gb recommended, but you could need more, for example Trinity uses ~ 1Gb/1M reads). In this case, it would be also recommendable more than 12 cores and a lot of hard disk space (at least 1 Tb). Once you have the reference the rest of the steps doesn't have big computational needs, except if you want to do a functional annotation of the reference using Blast in which case probably you will need a lot of cores (or a lot of time).

Best regards,

Yes. Workstation is ok for NGS analysis. Dell Precision is OK. But you'd better choose linux, big storage and big memory like 96G or more.

Yes you could go for Dell Precision T7600 it has all configuaration required for NGS analysis

I would focus on hardware specs rather than producer brands. To do so, first you have to identify your needs. How much storage will you need to keep your data online (hard drive capacity)? How big is each file you are going to analyze (RAM size and speed, hard drive speed)? Are the software you are going to use ready to use multiprocessors/multicore architectures (processor family)? Is execution time important (motherboard type, core frequency)?

I agree with Dr. Guarracino and will be more focused in the hardware specifications rather than the brand itself. If you are going to do RNAseq analysis the first question would be: Do you have a reference (for example Arabidopsis or rice) or do you need to do a de-novo assembly ? How much data do you have ? What tools do you want to use ? Do you have any previous knowledge of Linux and command line tools ? Here some orientation:

+ If you have a reference, you will not need a big server/workstation. Generally you can follow the protocol detailed in Trapnell C. et al. 2012 (http://www.nature.com/nprot/journal/v7/n3/full/nprot.2012.016.html). Bowtie2 needs ~2.2Gb to map reads to human genome size. More cores will speed up the mapping, but they are not strictly necessary. Tophat2 and Cufflinks are usually less demanding than Bowtie2 so I think that desktop with 8 or 16Gb should be enough in most of the cases. The downstream analysis using R could be a problem if you have a more than 3 conditions to compare. An alternative may be to use Galaxy and load the data to the Galaxy servers for the analysis (http://galaxyproject.org/) but again it will depend of how much data do you have.

+ If you don't have a reference, you will need to create one doing a de-novo transcriptome assembly using Trinity, ABySS-trans, SOAPdenovo-trans or any other assembler of your choice. In any case you will need a big server with at least 64 Gb of RAM memory (128 Gb recommended, but you could need more, for example Trinity uses ~ 1Gb/1M reads). In this case, it would be also recommendable more than 12 cores and a lot of hard disk space (at least 1 Tb). Once you have the reference the rest of the steps doesn't have big computational needs, except if you want to do a functional annotation of the reference using Blast in which case probably you will need a lot of cores (or a lot of time).

Best regards,

I do NGS analysis on my simple laptop with only 4GB RAM and everything works. The only thing you need to install Linux otherwise pretty much every computer might be used.

I did analysis of NGS on cluster consisting of 1400CPU, several terabites of HDD and 1200 Terabites RAM. The difference in analysis between the cluster and my laptor is 3-4 times not more :)

I also agree, brand is not important. Make sure to have:

i) plenty (!) of RAM (128GB is not a luxury!),

ii) fast multi-core processor (I think 4-6 core Intel i7 are the best currently),

iii) fast and redundant storage (RAID 1, 5 or 10).

Lastly, Linux is the must! Strongly recommend Ubuntu.

Thanks all for the valuable suggestions. Shall write again if I need more clarification.

Try out Dell Precision series, it should do the job for you (Try to have minimum 1 TB Space and atleast 8 GB RAM with a good processor and you can choose any brand)

go for hpZ820, upto 512gb ram, 2 -8 16 core processors, support 15krpm SAS hdd( max 4 hdd you can attach), may use lan card to achieve upto 10gbps speed ( giganet). can bur NAstorage of some good company. Search for good technical support of the said company. May also buy blades ...

@Bombarely :- out of Trinity, ABySS-trans, SOAPdenovo-trans which is best?

I mean it would be great if i get some help regarding this.

Can we follow BLAST for de novo assembly?. if yes then how appropriate/effective is that?

@Singh According some papers Trinity (I can remember right now the last that I read: Vijay N. et al (2012) Molecular Ecology DOI: 10.1111/mec.12014), but I think that if your really want to get the best assembly you should try at least two of them with different conditions (for example different Kmers). To evaluate how good or bad was the assembly you can take a look to the average size, but I like to see how similar is the length distribution to a close related species with a genome sequenced.

About BLAST, no I don't think that it is suitable for assemblies. Old assemblers such as TIGR assemblers used Smith-Waterman algorithm for assemblies (similar to BLAST) but with NGS will take to long (and al lot of memory) just to do the assembly doing a selfblast of the dataset and processing the matrix of results.

@Bombarely. Thank you . I will try out the same as you mentioned.