Abyss : Explain The K-Mer To Me!

Question

[madkitty]

Hey guys,

I ran a de novo assembly with ABYSS and I'm struggling to find the right value of k-mer. The average read length is 100bp and the insert size is 275bp. Here is a summary of k-mer / contig found that I had today :

K-mer    Contig Qty Smallest Contig    Biggest Contig
3         32            3                   3
4        136            4                   4
5        512            5                   5
6       2023            6                   6
7       6408            7                   9
8      28498            8                  11
9       9887            9                  35
10     4984           10                  63
15        132           15                2657
20         51           20                7205
22         41           25                9450
25         30           28                8402
28         24           31                9453
31         10           37               14686
50          5           68               16320
64          5           68               16334



n       n:200   n:N50  min    N80     N50     N20     max     sum
51      7       2      633    1618    2212    7205    7205    16230   /contig_k20.fa
41      5       1      589    1942    9450    9450    9450    16042   /contig_k22.fa
30      5       1      1022   2188    8402    8402    8402    16261   /contig_k25.fa
24      4       1      1616   2160    9453    9453    9453    16268   /contig_k28.fa
10      2       1      1616   14686   14686   14686   14686   16302   /contig_k31.fa
5       1       1      16320  16320   16320   16320   16320   16320   /contig_k50.fa
5       1       1      16334  16334   16334   16334   16334   16334   /contig_k64.fa

Can someone tell me to which k-mer value I should rely on ?

Answer 1

[maymay]

Something is either wrong with your data or the way you are using ABYSS. K-mer size of 10 is too short, you should not run assemblies with k below 19. These contigs with k-10 are probably mostly random. Anyway, what were you sequencing? Is it the whole genome/transcriptome or just a reduced fraction? For genome/transcriptome of Eukaryotes you want to get tens of thousands of contigs.

[madkitty]

We have a whole genome sequence of a mouse and I'm doing a de novo assembly of the mitochondria chromosome ChrM. Then which k-mer value should I take ?

Answer 2

[Shaun Jackman]

Hi,

Please report the assembly statistics as output by abyss-fac k*/*-contigs.fa

Cheers,
Shaun

[madkitty]

Hi Shaun, I have edited with the statistics, I'm not sure what you can do with that but anyway.. help is always appreciated :)