#23 · Genomics

Tree shape and scale constrain phylogenetic and recombining ancestral inference

The common thread is the computational cost of phylogenetic and ancestral inference as tree structure and data size limit analysis. subTreeLeap proposes an adaptive transition kernel that walks patristic paths to improve posterior tree-space mixing in Bayesian phylogenetic MCMC. On the recombination side, a tunable SMC(k) approximation to the coalescent, implemented in msprime, trades simulation speed against long-range ARG structure, and a benchmark of SINGER, Relate, and tsinfer shows accuracy declining as genealogical tree imbalance increases under transmission of reproductive success. BROOQS resolves blob trees into level-1 phylogenetic networks via quartet similarity and spectral ordering, scaling to thousands of taxa. The items do not share one algorithmic framework. What is new is explicit attention to how genealogical tree imbalance affects ARG accuracy.