Skip to content

Major Improvements

Metabuli v1.2.0

Improved sensitivity via spaced k-mers and reduced amino acid alphabet

  • Three layers of mismatch tolerance:
    1. Amino acid-level k-mer search allows for synonymous mutations (original feature).
    2. Reduced amino acid alphabet groups similar amino acids together, allowing for conservative substitutions (NEW).
    3. Spaced k-mers allow for mismatches at specific positions in the k-mer (NEW).

Alt text Figure 1. Robust homology detection via spaced metamer chaining. Alignment of diverged query and reference sequences demonstrating the combined use of reduced amino acid alphabets and spaced seeds. Identical residues are indicated by (|). Conservative substitutions (+) are tolerated through exact matching using reduced amino acid alphabets (e.g., F* grouping F, Y, and W). The spaced metamer's joker positions enables matching across amino acid mismatches at positions 4, 8, and 21. Overlapping metamers are merged into a single match chain, which is subsequently scored based on the underlying DNA Hamming distance (mutations highlighted in red) to provide a fine-grained, nucleotide-aware match score.

Improved scalability via syncmers

  • Syncmers are a subsampling technique that selects a subset of k-mers based on the presence of specific s-mers. This reduces the database size and classification time, minimally decreasing sensitivity in remote homology detection.
  • Two times smaller database and two times faster classification when c = (k - s + 1) / 2 = 2.