Fig. 5.

Benefit of the proposed minimizer sampling optimizations to long read mapping performance. We compare three methods: ‘Standard’ (default minimizer sampling, no masking), ‘Optimization 1’ (robust winnowing) and ‘Optimization 1 + 2’ (robust winnowing and weighted minimizer sampling). All three methods select at least one minimizer in each window. The left plot indicates total mapping time whereas the right plot indicates total memory usage, using datasets D1st. The x-axis is log-scaled in both plots. The combination of both optimizations is crucial in Winnowmap to achieve good efficiency