Using ABySS to assemble kmers into contigs

Amin Ghane

unread,

Jul 12, 2022, 1:53:53 PM7/12/22

to ABySS

Hi all,

I have been trying to use ABYSS to assemble kmers (31-mer) into larger contigs so that I can do a blast analysis to find a specific region of a genome but I keep getting an error which really does not give me any clue to find what is wrong with my code.

My input file format is like the following lines (fasta format and the id line is actually a p-value)

>2.975e-13
CTGTTTGTTTATGTCACACACAGACAAACAC
>2.621e-13
CTTCACCACACATGTGTGTATTGTGGAGTGA
>3.308e-13
CTTCAGATGGTATGCTCTCCAATTGATGTCA
... (+ 60 more kmers)

I am using a slurm HPC cluster so I have plenty of available memory!
The highlighted line in yellow is my code and the error is written in bold.

ABYSS -k31 -c0 -e0 input.txt -o output.txt

ABySS 2.0.2
ABYSS -k31 -c0 -e0 input.txt -o output.txt
Reading `input.txt'...
Loaded 63 k-mer
Unable to determine minimum k-mer coverage
Using a coverage threshold of 1...
The median k-mer coverage is 1
The reconstruction is 63
The k-mer coverage threshold is 1
Setting parameter E (erodeStrand) to 0
Generating adjacency
Added 20 edges.
Pruning tips shorter than 1 bp...
Pruned 45 k-mer in 45 tips.
Pruning tips shorter than 2 bp...
Pruned 14 k-mer in 14 tips.
Pruning tips shorter than 4 bp...
Pruned 4 k-mer in 2 tips.
Pruning tips shorter than 8 bp...
Pruning tips shorter than 16 bp...
Pruning tips shorter than 31 bp...
Pruned 61 tips in 5 rounds.
Popping bubbles
Removed 0 bubbles
Marked 0 edges of 0 ambiguous vertices.
Assembled 0 k-mer in 0 contigs.
error: no contigs assembled

Thanks in advance for your help.
Amin

Lauren Coombe

unread,

Jul 12, 2022, 2:55:36 PM7/12/22

to ABySS

Hello Amin,

A couple of questions and suggestions.

First, we recommend using the most up-to-date version of ABySS - looks like you're on ABySS 2.0.2, and the most recent version is 2.3.5.

Also, how did you generate the input k-mers file? It's possible that how it was generated means that there isn't sufficient k-1 overlap between the sequences, thus ABySS wouldn't be able to assemble anything. Can you input the reads instead of those kmers?

Thank you for your interest in ABySS!

Lauren

Amin Ghane

unread,

Jul 13, 2022, 11:17:51 AM7/13/22

to ABySS

Hi Lauren,

Thanks for your help.

Unfortunately, the newest version is not available on the cluster I am using for my analyses. I have to run it on my local machine.

I am using this pipeline in order to find sex-specific regions in the genome. I had about 100 WGS samples in the first place and based on the manual, I need to find k-mers associated with sex (found in only one sex, not in the other), so the input for ABySS is not the reads but k-mers.

What about e and c? Do you think I should increase these parameters?

Thanks again.

Amin

Lauren Coombe

unread,

Jul 22, 2022, 5:07:36 PM7/22/22

to ABySS

Hi Amin,

If you can't update to the newest version, do try using the de Bruijn graph mode of ABySS - we made that default in newer ABySS versions. More info here: https://github.com/bcgsc/abyss#assembling-using-a-paired-de-bruijn-graph

In that mode, we use the de Brujn graph to attempt to extend the input reads (k-mers in your case), so you may have more luck with that. Use the minimum k-mer coverage threshold (1). Do you know that you have k-1 overlaps between the filtered k-mers? That will be required to have a successful assembly using a de Bruijn graph-based assembler like ABySS.