Figure 2

DivA’s workflow. Using a sliding window approach, four parameters are calculated for every sequence in every window. If there are conserved sites at the edge of each window those are trimmed and the parameter values for the window are recalculated. The threshold values for each parameter are then calculated and used to classify each window in each sequence as very divergent (potentially non-homologous) or truly homologous. Sequences from overlapping windows classified as outlier are merged, and the final coordinates are provided in the output file. The user can also obtain a new alignment file where the outlier windows are masked.