Fig. 1

Four random examples of SEGREG recalibration. Bases are divided into groups based on the criteria described in the text, and in each group segmental regression is applied. The correspondence between raw and empirical phred scores is plotted as dots for bases of the same group and the breakpoint is chosen by the regression algorithm, with the two segments regressed to dots of the same color. If the dot (group) shown in the plot is based on less than 100 observations, the corresponding empirical base quality is imprecise and hence is not considered in the regression analysis