7
$\begingroup$

What are the pros and cons of the different basecallers in Oxford Nanopore Technology Sequencing?

I am about to start a MinION run on my laptop. What should I consider when choosing my basecaller? Can I let MinION do its sequencing and generate fast5 files and call the bases later? I think I might use Albacore. Metrichor and MinKNOW can also do basecalling. Are there reasons to choose one over another?

$\endgroup$
4
$\begingroup$

First of all - yes, you can generate FAST5 files and basecall later. Basecalling during the sequencing run is useful if you want results more quickly. You can also recall your FAST5 files with multiple basecallers, if desired.

There are several ways to basecall currently:

  1. MinKNOW
  2. Albacore
  3. Guppy

MinKNOW uses an embedded version of Albacore to perform its basecalling. However, the version in MinKNOW occasionally lags behind the released Albacore version.

Guppy is a GPU enhanced basecaller, and has not yet been released by ONT. It is used by default on the GridION. Results are similar, but not identical to Albacore, but Guppy is much faster if you have a supported GPU.

There used to be a cloud version of basecalling in Metrichor, but this is no longer available, and all basecalling must now be performed locally.

I'd suggest looking over Ryan Wick's comparison of the different available base callers, as he has done a comprehensive look (mainly from a qualitative perspective) at the different basecallers.

$\endgroup$
6
$\begingroup$

There are also third party free and open source basecallers that haven't been developed by Oxford Nanopore.

Of particular note is Chiron, which gave the best uncorrected assembly identity among the base callers that Ryan Wick tested. It's slower than Albacore, but appears to be more customisable, and could theoretically be used to model any DNA feature that is detectable in the electrical signal, rather than just the A/C/G/T base sequence.

There's a paper about Chiron on BiorXiv.

$\endgroup$
5
$\begingroup$

It's worth noting two things as of Dec 2018:

  1. Albacore is being deprecated (but is still available from the Nanopore developer portal).

  2. Guppy is under active development, so Ryan Wick's comparisons may not reflect the current state of things. (The claim is that the current, just-released version of Guppy uses a new "flip-flop" algorithm that improves accuracy over Albacore.)

  3. MinKnow contains the current version of the production basecaller. As of the release today, this has switched from Albacore to Guppy. If you want, you can run MinKnow for your basecalling (which will run Guppy), but you will likely need a beefy computer with access to a good amount of storage to do that.

$\endgroup$
1
$\begingroup$

Depending on your local system I'd suggest to run the MinION without local basecalling, and perform basecalling later using Albacore on server/cluster. That will keep storage requirements on your nanopore-computer low. Have albacore output fastq files which will be okay for most typical applications. Keep the fast5 files to repeat basecalling with newer versions of basecallers and to call nucleotide modifications.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.