2
$\begingroup$

NCBI has several labels for assembly completeness - Complete, Scaffold, Chromosome and Contig. Complete would be a circularized genome (or linear, rarely)

For a Complete genome it's fairly straightforward - if not present, a gene can safely assumed to be missing (or possibly present on plasmids, if those aren't available in the database).

But what about the others? With Contig-level genomes, I'd feel uncertain about making any statements (especially since some of these projects are split into thousands of contigs). But what about Scaffold, or Chromosome? Are they expected to be complete in terms of genes, but just missing repetitive sequences?

$\endgroup$
0

1 Answer 1

2
$\begingroup$

I would suggest that in the absence of "Complete" no very confident assertion can be made about gene presence/absence in a genome assembly. You need to actually have the chromosome sequence to make strong statements.

I think that most people (me included) would reluctantly accept that genomes with very good BUSCO or CheckM numbers (depending on your organism's taxonomy) are probably more or less accurate in terms of presence/absence. For metagenomics, I would extend that to the MIMAG criteria, which are pretty stringent.

These criteria are of course independent of the NCBI categories, which are pretty coarse-grained.

That said, I think that you can make stronger arguments about gene presence/absence from the sequencing reads from which the genome assembly is derived. Find all the reads for your genome, map them to the genome assembly and also to the closest related ortholog of the gene in question. How well is the gene covered? How do the alignments look? Harder for metagenomics, but there are still probably tricks one could use.

$\endgroup$
2
  • $\begingroup$ Thank you! Regarding your last paragraph, are you saying that it is probable that genes would be lost during assembly / binning (therefore not fully trusting a non-complete assembly), but not so much during sequencing (therefore trusting the raw data)? I imagine that might not hold for low sequencing depths, though. I was wondering, how high do you thing the sequencing depth have to be in order for the raw data to be reliable in terms of gene absence? $\endgroup$
    – Laura
    Jul 1, 2022 at 7:46
  • $\begingroup$ @Laura yes, I believe that's accurate. Many things are lost/misassembled during assembly that are obviously present in inspecting raw data. You would be looking for a similar coverage to the rest of the genome, unless you expect some funny business in that regard. The issues that arise are in the case of repetitive or highly homologous sequences, which is more of an issue in metagenomics as you might have separately binned closely related or homologous genomes. For example if you have abundant genome A, that will tend to suck up a lot of reads for homologous sequences in genome B. $\endgroup$ Jul 1, 2022 at 17:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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