Forum:PC build for Nanopore data analysis

Question

[stan.aanhane]

Hello,

From my department, we are looking at a PC for Nanopore sequencing. Unfortunately, because we are not entirely at home in the computer world, it is difficult for us to estimate what we need. That's why I'm asking for your help.

The nanopore sequencing itself has no special requirements. These are also neatly supplied by Oxford. We are building a pipeline for the data that is subsequently generated. The PC should also be able to handle this pipeline. We will make a pipeline to complete metagenome analysis to determine specific correlations between microbiome in different conditions. This will be done in python. All our question now is, what are the minimum/ideal specs of the PC. Could you help us with this?

If you have any additional questions, please let me know!

I really appreciate any help you can provide!

Best,

Stan

[andres.firrincieli]

An NVIDIA GPU is necessary if you want to complete the basecalling in a reasonable amount of time. Depending on the budget, any NVIDIA GPU with a cuda compute capacity of 6.1 or higher should works fine

[Jean-Karim Heriche]

An NVIDIA GPU is necessary

only if software can take advantage of it. Not every tool out there is GPU-capable.

[GenoMax]

ONT basecallers are performant only with GPU's.

[Jean-Karim Heriche]

I assumed that this was already covered because of the mention of

a pipeline for the data that is subsequently generated

[stan.aanhane]

Thank you, I am aware of this. As stated in the post, the requirements for the MinION are provided by oxford. Its mainly the amount of performance we need for the pipeline we are building, which would be mainly RAM and CPU power.

Answer 1

[GenoMax]

I assume you are referring to MinION sequencing since other larger ONT sequencers are supplied with built in compute for doing some of the initial basecalling etc. See this post for recommendations for hardware when using MinION.

Answer 2

[Jean-Karim Heriche]

It really depends on the details of your workflows and whether you'll need this computer for other things. You could look at papers describing similar workflows, some report the compute environment used for testing, e.g. the paper Assembly methods for nanopore-based metagenomic sequencing: a comparative study reports running all methods on a computer with CPU: AMD RYZEN 7 1700X 3.4GHZ; Cores: 8; Threads: 16; RAM: Corsair Vengeance 64 GB; SSD: Samsung 860 EVO Basic SSD 500 GB.
If you have access to an HPC this might be a preferable solution.

[stan.aanhane]

We are not done with the details yet. But the current steps are; Base-calling (guppy) -> classification -> Alignment control, and from there, we are still debating how we want to determine correlation. We expect large datasets of around 50Gigabases. The budget is relatively high, at about $10.000. We are currently looking into: AMD Threadripper 3970X (32 cores), 256 GB of ram, and an RTX 3090. But we want to be sure it's enough. But looking at literature, it should be sufficient.