Weighted syncmer sampling improves long-read mapping

Syncmers represent a novel class of methods for selecting k-mers that exhibit robustness against mutations in flanking sequences and demonstrate superior conservation in mutated sequences. Nevertheless, syncmers may generate a higher frequency of repetitive seed matches compared to alternative techniques, such as minimizers, which can result in increased computational time. In this article, we introduce weighted minimizer sampling, which integrates weighted minimizer sampling and syncmer sampling to enhance the sensitivity and accuracy of long-read mapping. We modified two state-of-the-art long-read mappers, Minimap2 and Winnowmap, by substituting the sketching sampling methods with weighted minimizer sampling. We assessed their sensitivity and accuracy using simulated and real datasets. The experimental results indicate that weighted minimizer sampling significantly improves the sensitivity and accuracy of long-read mapping. The source code is available at GitHub: https://github.com/hexinkaoan/weighted-syncmer.