Feature request: implement "circular aware" binning

[luispedro]

It would be great if the user could provide the list of circular contigs to avoid creation of spurious bins.
— https://twitter.com/fplazaonate/status/1603366542825119746

"Circular-aware" binning is a simple but important feature for long-read assemblies - circular contig MAG recovery can increase by 66%! We first noticed this issue in the hifiasm-meta paper, and @0xfxfxf describes other possible solutions here: https://arxiv.org/abs/2210.00098
— https://twitter.com/DPortik/status/1603412513915228161

[dportik]

Hi @luispedro,
Thanks for all of your work on semibin2!

I still think this would be a great feature to add, particularly for long-read assemblies.

Specifying some set of contigs to move to their own individual bins before binning the remaining contigs would be ideal. It's an essential step for any long-read binning workflow, and this would allow semibin2 to be used as a standalone tool (vs. its implementation in HiFi-MAG-Pipeline, for example).

We've since moved on to a "completeness-aware" strategy, which is similar but more robust:

Long contigs >500kb are individually examined using CheckM2 to determine percent completeness. All long contigs that are >93% complete are moved directly to the final MAG set, in their own individual bins.

Benchmarks and further description is here, in case it is of interest.