Confidence regions and hypothesis tests for topologies using generalized least squares

Area of Science:

• Phylogenetics and Evolutionary Biology
• Statistical Inference in Biology
• Computational Biology

Background:

• Confidence regions are crucial for hypothesis testing in phylogenetic studies, aiming to identify the true evolutionary tree.
• Current methods for constructing confidence regions in phylogenetics, such as the Shimodaira-Hasegawa and Swofford-Olsen-Waddell-Hillis tests, often yield conflicting results, being either too conservative or too restrictive.
• These discrepancies highlight the need for a more reliable and accurate approach to defining confidence regions for phylogenetic topologies.

Purpose of the Study:

• To develop a novel and computationally efficient method for constructing confidence regions for phylogenetic topologies.
• To address the limitations of existing methods that produce conflicting results in phylogenetic inference.
• To provide a broadly applicable statistical framework for evaluating evolutionary tree uncertainty.

Main Methods:

• A new confidence region for topologies is constructed using a generalized least squares (GLS) test statistic.
• The methodology is designed to be computationally inexpensive and applicable to maximum likelihood distances.
• The approach leverages statistical principles to ensure accurate coverage probabilities under specified model assumptions.

Main Results:

• The proposed generalized least squares method offers a computationally inexpensive and broadly applicable approach to constructing confidence regions for phylogenetic topologies.
• This method aims to provide more accurate and reliable confidence regions compared to existing, often conflicting, methods.
• Under the assumption of a correct underlying model, the method demonstrates correct coverage probabilities with a large number of sites.

Conclusions:

• The developed generalized least squares method offers a robust and efficient alternative for constructing confidence regions in phylogenetic studies.
• This approach mitigates the issues of over- or under-inclusion of topologies seen in previous methods.
• The methodology's computational efficiency and broad applicability make it a valuable tool for inferring evolutionary relationships with greater statistical confidence.