Thread by @rrwick

I have a little tool to share, in case it's useful to anybody else:
github.com/rrwick/Assembl…

Nothing too fancy, but if you work with large numbers of bacterial genomes, read on... (1/9)

I occasionally encounter a situation where I have lots of genome assemblies, but some are quite redundant. Often because the set contains outbreaks of near-identical genomes. (2/9)

For example, if you download all Klebsiella pneumoniae genomes, you'll get many thousands, but common disease-causing lineages (like CG258) are heavily represented. (3/9)

If you want to analyse the diversity in your genome assemblies (e.g. building a pan genome), this redundancy can be a pain. Near-identical genomes don't add much diversity but can make analyses slow. (4/9)

So in these cases, I usually want to reduce the assemblies to a unique subset. Using Klebsiella pneumoniae as an example again, I might want to reduce my set to a few hundred that are all quite distinct. (5/9)

That's where this tool comes in! It takes a directory of assemblies as input and clusters them, copying one assembly per cluster to an output directory. (6/9)

The representative for each cluster is chosen based on N50, so completed genomes are preferred over draft genomes. And the clustering threshold is of course adjustable, so you can get more or fewer output assemblies. (7/9)

It uses Mash (which is very fast, thanks @BrianOndov and @aphillippy!) to get the distances between assemblies, so it works on large input sets (10s of thousands of assemblies are fine if you have a decent amount of RAM). (8/9)

That's all! It's not a particularly sophisticated tool, but I hope it might be handy for someone out there. (9/9)