Open alexeigurevich opened 5 years ago
When the sequenced sample is quite similar to the reference with little structural variation and the quality of the reference is good, I also compare NGA50 to NG50 as a measure of the quality of the assembly. When these assumptions do not hold (incomplete reference or significant structural variation), I often find that the NGA50 will be much less than the NG50 (say 1 Mbp NGA50 but 20 Mbp NG50). When I look at a dot plot of the assembly compared to the reference however, the alignments are largely collinear. That is, there are no major translocations or chromosome-level misassemblies. I'd like to have an assembly quality metric that captures the large-scale correctness and contiguity of the assembly. So for example isn't interrupted by large scaffold gaps or inversions (which could reasonably be structural variation). That's not a very concrete proposal, just a brain storm. Have you observed a similar issue with NGA50 with very contiguous (megabase level) assemblies?
Nice suggestion from @aphillippy: add a new type of metrics measuring the ratio of NGA50 to NG50 (or vice versa; can be also applied to other similar metrics like NAx/Nx).