QuCo

QuCo: Likelihood-based co-estimation of species and gene trees

QuCo co-estimates species trees and gene trees under a likelihood-based framework by analyzing quartet subsets of taxa and jointly optimizing species tree topology, internal branch lengths, and gene tree posterior probabilities.

Key Features:

• Likelihood-Based Co-estimation: Simultaneously estimates species trees and gene trees, reducing biases associated with summary-based approaches such as long-branch attraction.
• Quartet Division Strategy: Decomposes taxa into quartets to reduce computational complexity and enable rapid likelihood calculations based on gene tree topologies while initially ignoring branch lengths.
• Marginalization Over Gene Tree Topologies: Computes the most likely species tree topology and internal branch length for each quartet by marginalizing over alternative gene tree topologies to mitigate gene tree estimation error.
• Posterior Probability Updates: Iteratively updates gene tree posterior probabilities conditioned on the inferred species tree to refine both gene and species tree estimates.

Scientific Applications:

• Phylogenomics: Infers evolutionary relationships among species from genome-scale data, improving robustness to gene tree errors and outperforming summary-based methods such as ASTRAL on estimated gene trees and empirical datasets.

Methodology:

QuCo takes independently inferred gene tree posterior distributions as input, partitions taxa into quartets, estimates the most likely species tree topology and internal branch length for each quartet by marginalizing over gene tree topologies, ignores branch lengths during initial likelihood computations to simplify calculations, and iteratively updates gene tree posterior probabilities based on the inferred species tree.