#35 · Genomics

Scalable and consistent methods for inferring species trees and phylogenetic networks

The dominant thread is the design of tractable phylogenetic inference methods under explicit statistical or optimization criteria. ASTRAL-X redesigns coalescent-based ASTRAL species tree inference to scale to 300,000 taxa with near-input-size memory while preserving statistical consistency. NetEC extends duplication episode clustering from trees to phylogenetic networks, pairing a polynomial-time dynamic programming feasibility test with an exact or heuristic optimizer. The third work introduces a semidefinite programming relaxation with iterative rounding for Balanced Minimum Evolution inference, validated on simulated and additional datasets. The shared contribution is computational tractability: ASTRAL-X addresses scale, NetEC addresses network-structured histories, and the semidefinite approach addresses a distance-based criterion. The episode clustering and Balanced Minimum Evolution papers are published, while ASTRAL-X is a preprint.