Evolutionary inference via the Poisson Indel Process.

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Joint statistical inference of phylogenetic trees and multiple sequence alignments is crucial for understanding molecular evolution.
• The classic Thorne, Kishino, and Felsenstein (TKF91) model, while foundational, presents computational intractability due to its exponential complexity in the number of taxa.
• This intractability hinders efficient phylogenetic analysis of large datasets.

Purpose of the Study:

• To develop a novel stochastic process for joint phylogenetic inference that overcomes the computational limitations of existing models.
• To introduce a computationally efficient alternative to the TKF91 model for phylogenetic and sequence alignment analysis.
• To enable scalable and accurate reconstruction of evolutionary histories.

Main Methods:

• We propose the Poisson Indel Process (PIP), a stochastic process related to the TKF91 model but with a different treatment of insertions.
• The PIP model is characterized as a Poisson process on the phylogeny, allowing for decoupled computations.
• Bayesian inference methods are employed to compare the PIP model with separate inference approaches.

Main Results:

• The PIP model reduces the computational complexity of joint phylogenetic inference from exponential to linear with respect to the number of taxa.
• This linear complexity facilitates significantly faster and more scalable phylogenetic analyses.
• Illustrative experiments demonstrate the effectiveness of Bayesian inference under the PIP model.

Conclusions:

• The Poisson Indel Process (PIP) offers a computationally tractable and efficient framework for joint phylogenetic tree and multiple sequence alignment inference.
• This advancement allows for more feasible analysis of large molecular sequence datasets.
• The PIP model represents a significant improvement over the TKF91 model in terms of computational performance.