Fig. 9

The optimal parameter settings estimated using Bayesian information criterion. The max depth correlates with sequence size (spearman correlation of 0.94). In contrast, the min count parameter does not correlate with sequence size (spearman correlation of 0.25), and every k-mer in the tree at the optimal depth occurs more frequently than any parameter we test. Therefore, we also include the frequency of the 5% least frequent k-mer in the tree. The x axis is log-scaled. The function fit is a logarithmic function of the sequence size \(y=c + \log (ax)\), with a standard error of 1.67 for the max depth