Fig. 3

Illustration of spaced-seeds, minimizer selection, syncmer selection and context dependency. Here, two sequences \(s_1\) and \(s_2\) are different from a single mutated base (second base, in pink). When comparing those sequences, one would like to focus on common bases, i.e., bases highlighted in grey. In the left panel, we present spaced-seeds for \(k=5\), with a wildcard at second position (represented by an underscore). We observe that identical spaced-seeds can be spelled over a mutated locus, e.g., T_GAGG. In the middle panel, we present a selected minimizer with \(k=5, w=3\). One blue window is presented, a second is suggested in lighter blue. A star shows the position of the selected k-mer in the window (we use lexicographic order). The mutated base has an impact on the overall window content, therefore a k-mer from the (unmutated) region of interest in \(s_1\) is no longer selected in \(s_2\). On the contrary, in the right panel, we show that syncmers can be more robust in this situation. We choose \(k=5, s=2\) and present closed syncmers. We underline the smallest \(s-\)mer in each k-mer in blue and a star shows the selected k-mers. We see that in this example, the mutated base has no impact on the syncmer selection, and the same syncmer is selected in the region of interest for \(s_1\) and \(s_2\)