Greetings everyone,
I’m doing my first genome assembly for a non-model plant species, and I need some insight about best kmer lengths for DBG-based assembly. My NGS data is a full lane of Illumina HiSeq V4 2x125 with a single library of insert size 350. Using kmer-counting methods, such as Jellyfish, my predicted best value k has ranged from 93-101 due to the large number of reads and their relatively-long length for HiSeq reads.
However, I have found all of my highest-quality assemblies at kmer-lengths <40, with my highest N50 and CEGs-mapped occurring at lengths 29 and 33, respectively. Does anyone know why I’m seeing such a large difference between predicted best k and actual best k? Given that it’s a discrepancy of more than 50bp, I feel like there’s got to be a common explanation that googling simply hasn’t turned up. I predict it’s related to heterogeneity in the reads, but I’m unable to find much elaboration on the effect of heterogeneous reads, so thanks for any insights!
I’m doing my first genome assembly for a non-model plant species, and I need some insight about best kmer lengths for DBG-based assembly. My NGS data is a full lane of Illumina HiSeq V4 2x125 with a single library of insert size 350. Using kmer-counting methods, such as Jellyfish, my predicted best value k has ranged from 93-101 due to the large number of reads and their relatively-long length for HiSeq reads.
However, I have found all of my highest-quality assemblies at kmer-lengths <40, with my highest N50 and CEGs-mapped occurring at lengths 29 and 33, respectively. Does anyone know why I’m seeing such a large difference between predicted best k and actual best k? Given that it’s a discrepancy of more than 50bp, I feel like there’s got to be a common explanation that googling simply hasn’t turned up. I predict it’s related to heterogeneity in the reads, but I’m unable to find much elaboration on the effect of heterogeneous reads, so thanks for any insights!