The Anti-Lexicographic SUS-anchor

Minimizer schemes Link to heading

Input: a window \(W\in \Sigma^{w+k-1}\)

Minimizer scheme: Sample the \(k\)-mer with the one with smallest hash: \[f(W) = \mathrm{argmin}_{0\leq i\lt w\ \ } h(W_{i\dots i+k}).\]

Sampling scheme: Arbitrary function \(f: \Sigma^{w+k-1}\to \{0, \dots, w-1\}.\)

Density: The expected fraction of sampled \(k\)-mers in a random string:

  • EADCAE.....
  • .ADCAEB....
  • ..DCAEBE...
  • ...CAEBEC..
  • ....AEBECD.
  • .....EBECDC
  • EADCAEBECDC

Density bounds Kille et al. 2024 Link to heading

Conjecture: Optimal schemes exist for \(k\equiv 1\pmod w\)?

Results Link to heading

Today: near-optimal selection schemes Link to heading

Selection scheme

  • \(k=1\), i.e., sample a position: \(f: \Sigma^w\to \{0, 1, \dots, w-1\}\).

Bidirectional anchors Loukides, Pissis, and Sweering 2023

  • Sample the start position of the lexicographically smallest rotation of \(W\).
  • EADCAE.....
  • .ADCAEB....
  • ..DCAEBE...
  • ...CAEBEC..
  • ....AEBECD.
  • .....EBECDC

Drawbacks of bd-anchors Link to heading

  • Rotations wrap around:
    • FBADC..
    • .BADCA.
    • ..ADCAE
  • Small strings cluster:
    • AAABCD....
    • .AABCDE.
    • ..ABCDEF

SUS-anchors Link to heading

  • Sample the smallest unique substring
  • CABBAB.
    • C A B B A B CA AB BB BA AB CAB ABB BBA BAB CABB ABBA BBAB CABBA ABBAB CABBAB
    • C A B B A B CA AB BB BA AB CAB ABB BBA BAB CABB ABBA BBAB CABBA ABBAB CABBAB
    • C A B B A B CA AB BB BA AB CAB ABB BBA BAB CABB ABBA BBAB CABBA ABBAB CABBAB
  • CABBAB.

  • Equivalent: consider the smallest unique suffix

  • SUS-anchors are forward
    • FBADC..
    • .BADCA.
    • ..ADCAE
  • but small strings still cluster:
    • AAABCD..
    • .AABCDE.
    • ..ABCDEF
  • Goal: Avoid overlapping SUS-anchors

The anti-lexicographic SUS-anchor Link to heading

  • Anti-lexicographic order:

    • compare the first character normally, and the remaining characters in reverse.
    • The smallest string is AZZZZ....
    • Like 10-minimizers/DNA-spacers (Shur, Tziony, and Orenstein 2026)
  • Anti-lexicographic SUS-anchor: anti-lex smallest unique substring.

  • Small strings do not overlap anymore:
    • AAABCD..
    • .AABCDE.
    • ..ABCDEF
  • Some intuition
    • SUS-anchors are optimal when sampled substrings never overlap
    • Bad case:
    • 10001000100.
    • .00010001001

Density results Link to heading

Results on real data Link to heading

Conclusion Link to heading

Anti-lex SUS-anchors…

  • are parameter-free,
  • have near-optimal density,
  • and can be streamed in linear time.

References Link to heading

References Link to heading

Groot Koerkamp, Ragnar, and Giulio Ermanno Pibiri. 2024. “The Mod-Minimizer: A Simple and Efficient Sampling Algorithm for Long $k$-Mers.” In Wabi 2024, 312:11:1–11:23. Lipics. https://doi.org/10.4230/LIPIcs.WABI.2024.11.
Kille, Bryce, Ragnar Groot Koerkamp, Drake McAdams, Alan Liu, and Todd J Treangen. 2024. “A near-tight Lower Bound on the Density of Forward Sampling Schemes.” Edited by Yann Ponty. Bioinformatics, December. https://doi.org/10.1093/bioinformatics/btae736.
Loukides, Grigorios, Solon P. Pissis, and Michelle Sweering. 2023. “Bidirectional String Anchors for Improved Text Indexing and Top-$k$ Similarity Search.” Ieee Transactions on Knowledge and Data Engineering 35 (11): 11093–111. https://doi.org/10.1109/tkde.2022.3231780.
Shur, Arseny, Ido Tziony, and Yaron Orenstein. 2026. “10-Minimizers: A Promising Class of Constant-Space Minimizers.” In WABI 2026, 390:5:1–5:20. Dagstuhl, Germany: Schloss Dagstuhl – Leibniz-Zentrum für Informatik. https://doi.org/10.4230/LIPIcs.WABI.2026.5.