Branch length heterogeneity leads to nonindependent branch length estimates and can decrease the efficiency of...

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Branch length estimates are crucial for phylogenetic inference using maximum-likelihood (ML) and minimum-evolution (ME) methods.
• Branch length estimates are not statistically independent under ML or ME, which can affect phylogenetic accuracy.

Purpose of the Study:

• To investigate how among-branch length heterogeneity (BLH) influences correlations among branch length estimates (BLEs) in phylogenetic inference.
• To assess the impact of BLH on the efficiency of ML and ME methods.

Main Methods:

• Simulations and analytical approaches were used to study correlations among BLEs under varying degrees of BLH.
• Maximum-likelihood (ML) was performed using the Jukes-Cantor model.
• Minimum-evolution (ME) employed ordinary least-squares (OLS) branch lengths with p-distances and Jukes-Cantor distances, with and without among-site rate heterogeneity.

Main Results:

• Increased BLH led to a higher frequency and magnitude of negative correlations among BLEs.
• The efficiency of both ML and ME methods decreased as BLH increased.
• The shape of the true tree, which influences BLH, is a critical factor in correctly recovering the topology.

Conclusions:

• Researchers must consider BLH, as branches of the same true length may have different probabilities of accurate reconstruction.
• Methods minimizing interdependencies of BLEs may reduce estimate variance and covariance, and improve the efficiency of model-based criteria.
• Further research is needed to explore methods for reducing non-independence of BLEs in OLS and ML.