2026-10-12T01:54:58Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33433045 2026-10-08T07:42:23Z category_24184 category_24319 item_type_3 month_year_10_2026

An order-indexed transport framework for microsatellite population distances kai liu (10109566) Bioinformatic methods development Genetics not elsewhere classified microsatellite genetic distance optimal transport Wasserstein distance genotyping error <p dir="ltr">Microsatellites, also called simple sequence repeats (SSRs), occupy an ordered repeat-count space, but population relationships can depend on how strongly a distance emphasizes large allele displacements. We developed an order-indexed framework that holds population allele distributions and locus weights fixed while varying displacement sensitivity across orders below, at, and above one. Fixed quantile matching and minimum-cost transport coincide for p ≥ 1 but can differ below one. Both untransformed costs and their 1/p-power representations were evaluated, with different metric guarantees on either side of p = 1. Across 300 known-history evaluations sharing 100 ancestry replicates, complete-tree recovery was 87.3% for first-order transport and 19.0% for the fourth-order Wasserstein distance. A quantile cost at p = 0.1 recovered 98.0% of trees, but its power representation recovered none and its scaled distance error exceeded the first-order value. An alternative mutation simulator changed the relative performance of quantile and optimal couplings. Finite designs and observation errors preserved several low-cost advantages while exposing losses caused by sparse panels and quality-dependent locus retention. Empirical order profiles changed nearest-neighbor decisions in fish, bird, and insect comparisons; lower orders most consistently improved bird agreement with single-nucleotide polymorphism matrices, whereas insect nearest-neighbor agreement remained highest at larger orders. Increasing order concentrated transport on fewer loci, and family residuals showed different order responses for cost excess and tail probability. The complete profile therefore describes a task-dependent sensitivity spectrum: first-order transport supplies a common reference, lower orders compress displacement contrasts, and higher orders reveal population relations and marker diagnostics driven by large repeat shifts.</p> 2026-10-08T07:42:23Z Dataset Dataset 10.6084/m9.figshare.33433045.v3 https://figshare.com/articles/dataset/An_order-indexed_Wasserstein_framework_for_microsatellite_population_distances/33433045 CC BY 4.0