BQSR errors vs real variation
When I have a known rate of 1 SNP every 30 basepairs and my reads (only counting the 'good' ones, used in BQSR) have 1e10 basepairs in total. So, I have
1e10 / 30 = 3.3e8 SNPs on my reads.
Can I say that I expect BQSR to find
3.3e8 + epsilon errors when not masking any known SNPs?
This is connected to creating an own set of known SNPs for BQSR. I so far have the feeling from the reports, that the error rate is too high and my set should be bigger. Using the known rate of SNPs in my data as comparison to the Observations/Errors in the report and table from recalibration. So having the optimal set, I would get
3.3e8 + epsilon Observations and
Am I right?