data depth

[Wanjie-Feng]

A really great tool ！
I have a doubt: Will the depth of WGS or hifi data affect the final assessment result? What is the recommended minimum depth for WGS and hifi data?

[arangrhie]

Hi @Wanjie-Feng ,

Yes the results vary depending on the coverage. Low coverage (<10x) sometimes have regions with no sequencing reads, which could lead to no kmer - and if that region is present in the assembly, it could be picked up as a false error kmer lowering the QV.

My recommendation is 50x WGS or HiFi if possible, as a data scientist, more data always helps. This is a recommendation considering genome assembly, evaluation and polishing all together.

For evaluation only, you could go lower, from experience >20x (>10x per haplotype) was ok. The lower the coverage goes, the frequency overlap between error kmers and the 1-copy kmers start to increase, making it difficult to visually decipher in the spectra-cn plots. As long as the 1-copy peak is visibly distinguishable on the kmer histogram, I'd take whatever coverage it is. My best practice is to feed read.hist as input to genomescope2 (or something similar) and check for the kmer frequency histogram, ploidy and chances for contamination before beginning the assembly / evaluation.

meryl count k=31 output read.meryl reads.fq.gz
meryl histogram read.meryl > read.hist

Best,
Arang