Empirical assessment of sequencing errors for high throughput pyrosequencing data.

With Newbler, although the substitution error rate is very low in absolute terms, if we regard the number of substitutions relative to the overall number of errors (including indels), we have that they account for about 50% of the total number of errors in this data set (665969/(665969+389063+287991)), which is quite significant, although in comparison this is a lower proportion than what is observed in other technologies [ 21 ].

also in 5,695 other papers