workstation purchase for metadata analysis and running guppy

Question

[tanusaroha]

hello, i am a biologist with no background in computer science. i am planning to buy a workstation for metadata analysis like nanopore data assembly etc. i want to ask for which configuration should i buy and also i want to run a tool,named guppy, for that tool which graphic card (pre installed on workstation) would do the work? also, does nvidia tesla v100 comes as pre installed on any workstation?

Answer 1

[colindaven]

Why not buy the Minit instead ? That is our strategy and will save me a lot of time coordinating basecalling if it's all done by the device.

As a bioinformatician I can say that if you're not working on neural networks, eg Tensorflow etc, then GPUs are not that useful in bioinformatics. There are some accelerated approaches available, eg for nanopore assembly polishing, but these tend to be technically quite demanding (i.e. installing CUDA on Linux, configuring command line apps etc).

That said, having one graphics card and the skills to use it is going to be far more flexible than just a Minit.

[GenoMax]

MinIT recommendation would be applicable only if OP is planning to run single MinION. Would it be usable for other types of nanopore data?

[colindaven]

Gridion and Promethion come with basecallers as part of their hardware package.

Flongles run on Minion so can use Minit. Not sure other types exist yet ?

[prishly]

Guppy on GPU vs guppy on CPU is hours vs days of basecalling even for moderately large datasets. By the way guppy basecaller does use neural networks according to this, as do all the latest nanopore basecallers.

[colindaven]

Yes, guppy uses NNs. I did not say that it didn't :-)

Answer 2

[prishly]

Check out Mk1C

[GenoMax]

OP could be referring to basecalling offline with GridION or other kinds of higher throughput data. MinIT and Mk1C are specifically targeted MinION based on what I see on the webpage.