Raw error rate calculation:

[JohnK]

Hi, I've visited some previous threads that talked about error rates and filtering reads based on error rates. How does one go about calculating this error rate off of the raw sequence coming off the machine? I thought it might be the number of bases with errors over the total number of bases, or perhaps it's the average probability of error across all reads? If someone could give me a general idea, it'd be very helpful. Thanks!

J

[csoong]

that is in general the correct method to calculate empircal error rate.
I believe samtools protocol has talked about this in the section utilizing GATK to recalibrate sequence run qualities.
Basically, since there are so many known snps that are counted as mismatch when aligned to the reference, one has to figure out a way to not count those known snps as errors.

for detail, read GATK manual section on recalibrating sequence qualities.