Recommended open-source assembler on desktop computer?

Question

[grantregis]

I'm new to bioinformatics and I'm trying to assemble many raw sequence reads from the SRA. (Sequencing done with Illumina HiSeq 2500). Most of the files contain between 115 M and 440 M bases, but some of the files have counts around 1.1 G bases. I know how to convert the data into formats such as .fastq and the like. I've looked around and found many assemblers, like SPAdes and Velvet. Can someone recommend an assembler can work quickly and effectively on a desktop computer (with 16 gb RAM)?

Thanks,

Grant

[surovisaikia]

What should be the configuration of a desktop for basic bioinformatics works like molecular dynamics. Please let me know what amount of RAM and graphic cards should be installed?

[ATpoint]

Hi surovisaikia, welcome to Biostars. We generally discourage posting questions in existing questions. Please use the search function to go through existing similar questions and if this still does not yield the desired answers, open a new question, providing the necessary details on what exactly you are working on and (in this case) what your budget would be.

Answer 1

[h.mon]

It will depend on which taxa you want to assembly - and on particular characteristics of each genome - but with such small memory, Minia or MEGAHIT are probably two good choices.

[ATpoint]

Just out of interest for me who never assembled a genome, what is necessary for assembling a human genome with Illumina short reads? Would a "normal" HPC node (64 Xeon cores, 128GB RAM) do or do you really need to go big on this?

[GenoMax]

Human genome de novo assembly may require 1TB+ (or even more) RAM (depending on the amount of data you have).

[h.mon]

Also bad quality reads increase memory requirements a lot. There are techniques to lower memory requirements, such as digital normalization and error correcting the reads.

[ATpoint]

Puh, that is a lot. Are the standard tools able to use cummulative memory from multiple nodes?

[colindaven]

In my experience, no. Fat nodes with lots of RAM are required. I wouldn't generally attempt vertebrate short read assembly on a machine with lower than 128 GB RAM, better 512 GB.

[h.mon]

Depends on the assembler, among other things, but if I recall correctly, Minia can handle a human genome with such configuration.