Open Syncmer?

[jianshu93]

Hi Team,

Just wonder whether open syncmer could be another option (https://github.com/bluenote-1577/local-kmer-selection-results/blob/db4c5c686433fb09f8571da7ca11b355aaa28e30/src/seeding_methods.rs#L45). It was not bit packed there but only a sample reference implementation.

Thanks,

Jianshu

[imartayan]

Hi @jianshu93,

From what I understand, this corresponds to selecting windows where the minimizer is at position t, am I correct?
Should we treat differently the case where a single minimizer appears at multiple positions? Or should we only consider the leftmost position?

[jianshu93]

This is my understanding and the local kmer theory paper explain it very well, for any kmer, the smallest s mer (s<k) by some order equal to some position t in the kmer, this kmer will be selected, otherwise not selected. In practice, t is alway the middle position of the kmer. see figure 1 in this paper(Theory of local k-mer selection with applications to long-read alignment).

For repetitive kmers in a sequence, all will be selected since those kmers are independent, so it is somehow slower than minimizer, but offer more information for sequence with more mutations than minimizer.

No window is needed, but scan all kmers in a sequence.

Thanks,

Jianshu