Minia bug with "--debloom original"

Question

[shammond]

Hi Rayan/Devon,

I was assembling a set of 2x250 human reads with 72x coverage using the beta version of minia 3 that you kindly provided me, when I encountered this error:

[bug detected in adjacency data structure] was supposed to delete node AAAAAAAAAAAAAAAAAAAAAGGTCTGGGGTGTCATTATATTGCCCAGGCTGGTCTCAAACTCTTGGGCTCAAGTGATCCTCCCACCTTGGCCTCATAAGAGTGTTGGGATTACAGGCGTGAGCCACTGTGCCCGCCCTGTCTCTATTTATACAAATAAA but couldn't find it :(.
assembly may be inaccurate around this kmer. will continue assembling anyway

It occurs (with different nodes) 3822419 times at k 144 and 3682804 times at k 160, but never at k 128 or below. The only difference between these runs, aside from k, is that I used "--debloom original" at k > 128.

Perhaps because of this bug, the run time has increased from ~20 h to > 100 h, and the N50 has dropped from 5777 (k128) to 822 (k144).

Is there a fix for this bug?

Thanks, Austin

Answer 1

[Rayan Chikhi]

No fix yet, I don't think I've tested that latest minia 3 on large genomes with k > 128 yet :) Thanks for pointing it out! Will let you know when a fix is out.