Hon - Collision Cb The Extra Match

We present Collision Cb, a novel approach to resolving extra-match ambiguities in sequence alignment for homology detection. Collision Cb integrates collision-based scoring with context-aware block (Cb) modeling to improve sensitivity in detecting distant homologs while reducing false positives from repetitive or low-complexity regions. Experiments on benchmark protein and nucleotide datasets show consistent gains in recall at fixed precision compared with standard pairwise alignment and profile HMM methods. We analyze algorithmic complexity, parameter sensitivity, and failure modes, and discuss applicability to large-scale database searches.

3.1 Overview Collision Cb proceeds in three stages:

3.2 Context-aware block (Cb) scoring For each match m, define a surrounding block B_s(m) in S and B_t(m) in T of size L (default 256–1024). Compute: Collision Cb The Extra Match Hon

3.3 Collision modeling and scoring adjustment Define adjusted score: w(m) = σ(m) · c(m) · (1 − γ · CollDensity(m)), where CollDensity(m) measures local collision density in G (normalized count of conflicting candidates) and γ∈[0,1] controls penalty strength. CollDensity is computed by considering proximity and score closeness to neighbors.

3.4 Selection algorithm

Key findings (summary):

Include representative tables and plots (recall vs precision, ablation bars, runtime comparisons). We present Collision Cb, a novel approach to

In double-elimination brackets, the Grand Finals sometimes have a “bracket reset” – an extra match if the loser’s bracket champion wins the first set. That extra match is often called the “final final” or “extra game.” In Hon, where matches could last 45+ minutes, an “extra match” was exhausting and legendary.