Computer requirements for RNAseq alignment

Question

[n.lu]

Hi,

I plan on analysing the following RNA seq data: cca 30M reads/sample, 15 samples, human data. The aim is to do differential expression analysis. I plan to perform alignment with HISAT2 and I wonder if my computer (specifications below) suitable for this?

• Processor: Intel(R) Core(TM) i7-4790S CPU @ 3.20GHz, 3201 Mhz, 4 Core(s), 8 Logical Processor(s)
• Installed Physical Memory (RAM) 16,0 GB
• HDD: WD Blue PC Desktop Hard Drive (WDC WD10EZEX) - 1TB
• SSD: Crucial BX100 SSD (CT1000BX100SSD1) - 1TB

If not I, what would my best option be - additional RAM (how much?) on this computer, different alignment tool, cloud computing (like AWS), galaxy?

Any input or experience is helpful! Thanks, Nina

[Bastien Hervé]

AFAIK, HISAT2 is the best alignment tool for low RAM. 1 TB is way enought to stock your data and output data. I would say, give it a try and if the alignment step failed, look at the output error to know if it is a memory issue or something else. If it is a memory issus you can try to add more RAM or run your alignment under a cluster. I have no experience with galaxy.

Answer 1

[Nicolas Rosewick]

You could use kallisto : https://pachterlab.github.io/kallisto/about . It uses minimal RAM with similar sensitivity as more "standard" "align-then-count" methods. In general less than 4Gb RAM per sample is required

From kallisto website :

On benchmarks with standard RNA-Seq data, kallisto can quantify 30 million human reads in less than 3 minutes on a Mac desktop computer using only the read sequences and a transcriptome index that itself takes less than 10 minutes to build.

Answer 2

[GenoMax]

For human genome you need ~30G of RAM for STAR/BBmap (if you are staying with alignments). So it would be advisable to get at least 32G. If you are able to use mapping type methods (noted by @Nicolas) then you could use your present config.

[Nicolas Rosewick]

it would better say pseudo-mapping or pseudo-alignment tha mapping methods ;)