Skip to content

v1.5.0 — uniqsketch --min-margin signature hardening

Latest

Choose a tag to compare

@hamidamz hamidamz released this 17 Jul 17:37

Features

  • uniqsketch --min-margin=N: harden signature selection against single-base sequencing errors.
    With --min-margin=2, uniqsketch prefers signatures that lie at Hamming distance ≥ 2 from every
    k-mer in other references, so a single substitution error in a read cannot turn a foreign k-mer into
    one of a reference's signatures — a source of false-positive hits. It is a tiered preference: within
    each genomic selection slot a 2-safe, non-low-complexity candidate is chosen when available, falling
    back to the previous behavior only when a slot has no 2-safe option, so per-reference signature counts
    and genomic spacing are preserved. A summary line reports how many selected signatures met the margin.
    The default (--min-margin=1) leaves selection unchanged. Currently supports margins 1 and 2; higher
    margins are planned.

Full changelog: https://github.com/amazon-science/uniqsketch/blob/main/CHANGELOG.md