Segminator II now contains a feature for the probabilistic read error detection across temporally obtained reads (PREDATOR).

In brief, the approach utilizes joint nucleotide frequencies across temporally sampled data sets from a single host, in conjunction with known platform error rates, to derive a set of posterior probabilities for each of the possible nucleotides at each site. Posterior values indicate the probability of a nucleotide observed within any of the individual data sets, regardless of frequency, being biologically relevant to the viral population and, thus, not being due to platform error. Error correction then involved the replacement of individual nucleotides associated with low posterior probabilities with the population consensus; thus avoiding the need to discard poor quality reads.


Figure 1:Predator within Segminator II

The sole requirement for the application of PREDATOR to the data is that more than one dataset is associated with the project. When this is the case the boxed area in figure 1 becomes active and the user can select a correction posterior. Once a posterior threshold has been selected all reads, trees and base frequencies within the software will reflect correction based on the value chosen. To undo the correctional procedure the slider can simply be set to 0 - indicating that all variants with a posterior of 0 or more will be used.