SIFT: Syncmer-Informed Selection of Haplotypes for T2T Imputation

SIFT enables phrase-level selection of reference haplotypes for genome patching by using closed-syncmer parsing plus pseudo-MEM queries, avoiding gap storage via parse equivalence.

• Why it matters: As pangenome panels grow, efficiently finding suitable reference haplotypes for genome patching becomes increasingly challenging. SIFT targets this bottleneck for ImpuT2T-style patching.
• What they did: SIFT parses reference and query sequences using closed syncmers, builds phrase-level indexes with run-length compressed suffix arrays, and adapts Boyer-Moore-Li to find pseudo-MEMs. It aggregates matching evidence to propose candidate haplotypes.
• The result: By showing that two closed syncmers fix the maximum consecutive distance in any string, SIFT ensures equal substrings across parses represent identical sequences without storing gap sequences. This provides a scalable basis for querying large reference panels at the phrase level.

AI summary of the abstract · may contain errors · not medical advice · read the paper.