Is this a complete bacterial genome? (CheckM)

Question

I've been working with a bacterial genome assembly. The initial stats look pretty good (3 main pieces, depth ~50x [20x - 100x] if I calculated it right). I'd expect this to be a virtually complete genome apart from some low-complexity regions.

However, when I use CheckM to estimate completeness / redundancy, I get ~99% completion, <2% redundancy. (out of 579 expected single-copy marker genes, 563 were present once, and 11 were present twice). I am especially concerned about the missing genes.

I can think of several possibilities: a) Limits of CheckM. This software uses a sample of 82 genomes to select marker genes for this phylum: however, there are no available genomes in the same genus as the one I'm working with. It might be possible that this is natural variation.

b) The "single-copy" genes present twice might be due to misassembly. But while the sample was not axenic, the organism of interest was majoritary, and long reads were used as well, so this shouldn't be an issue?

c) The genes were somehow missed by sequencing. (but the depth is pretty good?)

I don't know how likely a) is, so it's difficult for me to assess whether what I have is a normal result, or something that requires additional sequencing in order to make sense. I would appreciate your thoughts!

Answer 1

See this table from the CheckM table to put this number in context. In general, your numbers sound excellent, even compared to isolate genomes.

I'd recommend for context taking some finished microbial genomes (E. coli K12 etc) and running CheckM. I bet you'd be surprised what you find.

I think that you can probably not worry about (c) as a possibility. If you are very worried about (b), you could map reads against your assembly, filter out the ones that don't map well, and assemble those to see what you find. Long reads suggest that your genome is in good shape. You can of course do manual curation of your genome using tools like Anvi'o, or by scaffolding/gap-filling your assembly with long reads or an orthogonal strategy like optical mapping or Hi-C.

I think that the influence of (a) is significant. Overall it sounds like you have a reasonable genome.

Answer 2

Don't read into the CheckM result too much. Genes having one copy in some species may have multiple copies in other species and genes present in some species may be absent in others. This may even happen between strains, let alone between species or between genera. It is rare to see 100% checkM completeness and 0% contamination. A common standard for "near complete" assembly is >90% completeness and <5% contamination. Your assembly is far above this standard.

Meanwhile, beware that checkM only evaluates single-copy genes but multi-copy genes are usually harder to assemble. There might be complex multi-copy genes in the three assembly gaps. Without getting a circular contig, it is hard to say for sure your assembly is really complete.