Flye assembly, increased coverage results in decreased N50

Question

[roger.huerlimann]

Hi all,

We are sequencing a fish species using ONT long reads. We have assembled the genome using Flye.

The first assembly we did with about 22X coverage, and we got the following metrics:

No. contigs: 1,229

Largest contig: 28,606,363

Total length: 655,408,839

N50: 12,108,055

BUSCO Complete (%): 85.8

We then did two more minion runs, which were not perfect in terms of yield, but gave us better N50 distribution and we did an assembly with 32X coverage. While most metrics stayed the same, the N50 actually decreased by 1/4. Does anyone have an explanation why this could be happening?

No. contigs: 1,340

Largest contig: 27,194,123

Total length: 658,156,384

N50: 9,408,483

BUSCO Complete (%): 87.6

[samuel.a.odonnell]

Sometimes adding too much coverage doesnt improve the assembly but 32X isnt exactly exessive... \ Perhaps downsample to 20X with just the longest reads and try again?

[roger.huerlimann]

Hi Samuel,

Cool thanks, I will look into that and see how things change.

Cheers, Roger

Answer 1

[benformatics]

You may have resolved some repetitive region that split a large contig into two (or more). Hard to say if you want to know for sure just plot the lengths of contigs and see which ones changed dramatically (and investigate further), also check the other N90, N10, N75 etc...

I wouldn't really worry about the N50... it might just be that your contigs length were always hovering on the edge of that 50% of the genome mark. Also you increased the genome size by 3 million, maybe previously the 12M bp contig was like right at (655/2)M, now you need the next smallest contig (the 9M one) to make up for that extra 3M.

[roger.huerlimann]

Thanks!

Yeah, after some thinking I came to the conclusion that the assembly is probably still good enough, especially since we now will go into HiC scaffolding. Even an N50 of 9 Mbp is more than sufficient. That's also a good point about the 3 M size increase and the shift from 12 M to 9 M in N50!

Cheers,

Roger