Closed babarlelephant closed 2 years ago
For the coverage plot that I got
quite different to yours
Pulling up the base counts for that sample and visualizing them in SamBamViz, that region is extremely low-coverage (ranging between 2 to 4 reads per site):
Note that the SamBamViz visualization uses 0-based indexing, not 1-based
Regarding the differences in coverage plots:
(1) Are you using the trimmed BAM rather than the raw BAM? Your Pastebin says "niemasd merged BAM then samtools consensus", which makes me think that you're using the untrimmed merged BAMs rather than the trimmed BAMs. You should be using the trimmed BAMs.
(2) I'm using a minimum base quality score of 20 when I call SamBamViz to generate the plot, but you're not filtering by quality score as far as I can tell. In general, low-quality base calls need to be ignored in all steps of the analysis (I use a threshold of 20 as per the iVar recommendations)
Yes I'm using the merged untrimmed bams, which gives a reasonable coverage and sequence, you are perfectly right that filtering for quality needs to be done, but your filtering step may be too stringent at least for WH02.
Thank you!
but your filtering step may be too stringent at least for WH02
I'm not sure what you mean by this. A base quality score of 20 means 99% accuracy, and trimming/ignoring base calls with quality <20 is standard not only in viral genomics, but in most (all?) sequencing contexts:
If the coverage drops below a reasonable level after using a base call quality threshold of 20, "This might be too strict" is not the correct interpretation: the correct interpretation is "This sample has too few high-quality base calls covering these positions" (which is precisely what N
in the consensus sequence denotes).
Hi, this is what I obtained by downloading your merged bams and doing
samtools consensus -m bayesian --use-qual --call-fract 0.75 -a --min-depth 10 bamfile.bam -o seq.fasta --show-del yes --show-ins no
Any guess on why it gives such a different result for WH02? Your consensus sequence has 27512 N
The fasta I got: https://wtools.io/paste-code/bC6N
mutations wrt Wuhan-Hu-1