Maximum Likelihood Estimation of Symmetric Group-Based Models via Numerical Algebraic Geometry.

Area of Science:

• Computational Biology
• Phylogenetics
• Algebraic Geometry

Background:

• Phylogenetic models can be parameterized using polynomial maps related to root distribution and edge transition probabilities.
• Previous work established polynomial inequalities characterizing joint probabilities for symmetric continuous-time group-based models.

Purpose of the Study:

• To apply numerical algebraic geometry to maximum likelihood estimation in phylogenetic models.
• To explore scenarios where maximum likelihood estimates may not exist, particularly in complex phylogenetic models.

Main Methods:

• Utilized polynomial parametrization maps derived from phylogenetic models.
• Employed numerical algebraic geometry techniques for maximum likelihood estimation.
• Investigated specific examples to identify limitations of estimation methods.

Main Results:

• Demonstrated the application of algebraic methods to maximum likelihood estimation in phylogenetics.
• Identified a specific case where the maximum likelihood estimate for a phylogenetic model does not exist.
• Highlighted the difficulty of discovering such non-existence issues using traditional methods.

Conclusions:

• Numerical algebraic geometry provides a powerful framework for analyzing phylogenetic models and their statistical properties.
• Algebraic methods are crucial for uncovering edge cases and limitations in maximum likelihood estimation for phylogenetics.
• The study underscores the importance of advanced mathematical tools in computational biology for robust model analysis.