search

XLuyu / PDR

1 stars 1 forks source link

reference genome #2

Open Karimi-81 opened 3 years ago

Karimi-81 commented 3 years ago

Hi There, Thank you for developing the PDR. I have a general question in regard to reference and assembly genomes. Reasonably, the new draft of assembly should be more complete or even larger (size) than the previous draft. However, we usually use the previous draft as reference for comparison. For instance, in my case, the assembly size is of 2.6 Gb and is of high N50 (30 Mb) while the previous draft (reference) is so fragmented and has size of 2.4 Gb. Do you think the PDR is still a good measure to compare such situations? How we interpret for example the following results: Genome payload: 2.274143302E9 PDR Total: 4.5750723042110915E18 PDR Ratio: 0.8846313105144036

Thank you.

XLuyu commented 3 years ago

Hi,

Thank you for trying PDR. The motivation of PDR is to give an ultimate single number that can help people compare several assemblies and know which one is as informative as the reference genome, rather than comparing 2 drafts in terms of synteny.

For synteny analysis, an alignment plot (by Mummer) can be a better choice when drafts have only a few contigs (e.g. <50). However, if both drafts have quite a lot of contigs (e.g. >1000), the alignment plot may look like random dots and thus be uninformative. In this case, PDR can still help.

After all, the PDR ratio roughly indicates the percentage of information from the reference which is also in the assembly. For example, you got a PDR Ratio = 0.8846. This roughly means we found, in the new draft, 88.46% of information from the old draft.

Please note PDR is not symmetric. When a draft A is actually a superset of another draft B, PDR(A against B)=1.00 but PDR(B against A)=0.8846. So, I suggest that you can run PDR once again but swap reference and assembly. If PDR(A against B) is much higher than PDR(A against B), B is likely included in A. So, you may want to use draft A in downstream analysis.

Another choice is to use a close species genome as the reference. But this doesn't work all the time, subject to how close the species is. If you get PDR Ratio < 0.2, this doesn't provide reliable insight.

Luyu