meta-velvet returns nodes instead of contigs in assembly?

[deprekate]

Hi I just started using meta-velvet to assemble a small metagenome 100k. The average length of reads are 102, with a min of 29 and a max of 227. I tried running meta-velvet with a k-mer of 21.

After the assembly is finished and I look at the output file "meta-velvetg.contigs.fa" I noticed that there are contigs of length 21 (less then the smallest read). This behavior seems strange, that it is returning just the k-mer overlap (or node) and the rest of the reads did not match, and shouldn't be reported as a contig

[cliffbeall]

I am not sure if this applies to meta-velvet, I only have used the original. However, with standard velvet, it reports the contig lengths in kmers rather than nt in the fasta headers. So is it possible that you have contigs with 21 kmers, that are actually 41 nt long?

[deprekate]

Ah, yes you are correct. The length of the actual sequence is >41 even though the header says length_21.