BUSCO is a tool that assesses genome assembly completeness by finding a particular set of near-universal single copy orthologs. I am wondering, is there a general consensus on what score people would take to be a good one? I.e. should I ignore any assembly that scores less than (for example) 75% completeness?
I'd be interested to hear what people who (might) use this software would deem an acceptable score.
I think relative BUSCO comparisons are more informative than absolute BUSCO scores. For example, using BUSCO to compare various assemblies of the same data.
If you have the computation resources, you can also perform a saturation curve of input reads to BUSCO scores. For example, take random subset of 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% of your reads and perform an assembly. Then determine BUSCO score for each subset assembly and see if there is a plateau. This will tell you whether you have saturated all the genic information with the libraries you have. Perhaps a different library prep method is needed to get more diverse reads to improve your assembly.
Not sure if there exist any consensus here , the BUSCO score is what it is. Especially since there exists many different (sub)sets for this BUSCO, so I can easily get a very low score if I use the "wrong" set
In an ideal case it should of course be as close to 100% as possible
It is a bit difficult to say in an absolute sense what would constitute a good BUSCO score, but BUSCO is very useful for comparing different assemblies. If your best assembly scores less than 75% completeness, then you shouldn't ignore it, as it is better than having no assembly at all. Anyway, I would say an eukaryotic assembly with >85-90% single-copy BUSCOs is of good quality, a prokaryotic assembly should be >95% single-copy BUSCOs.
If you look at the first BUSCO paper supplementary material, you will see the wide range of BUSCO completeness for different lineages, and also the relation between BUSCO and N50.
