Indel information eliminates trivial sequence alignment in maximum likelihood phylogenetic analysis.

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Maximum-likelihood (ML) methods are widely used for phylogenetic tree estimation based on fixed sequence alignments (MSA).
• The incorporation of insertion and deletion (indel) information into these traditional ML frameworks remains underexplored.
• Standard ML approaches often treat indels as missing data, potentially overlooking crucial evolutionary signals.

Purpose of the Study:

• To investigate the impact of indel information on ML-based phylogenetic tree estimation using fixed MSAs.
• To determine the optimal alignment strategy under different assumptions regarding indel treatment.
• To evaluate the validity of current ML software implementations that assume indels as missing data.

Main Methods:

• Theoretical modeling using a single character example to illustrate alignment optimality under different indel assumptions.
• Empirical analysis of five diverse multiple sequence alignment datasets to validate theoretical findings.
• Comparison of likelihood scores between alignments treating indels as missing versus incorporating them into the character alphabet.

Main Results:

• A trivial alignment, different from that optimal for parsimony, is found to be optimal in ML when indels are treated as missing data.
• This optimality of the trivial alignment is not observed when indels are incorporated into the character alphabet.
• Likelihood analyses across five datasets consistently showed that the trivial alignment's optimality is dependent on the indel treatment.

Conclusions:

• Indel information is crucial for accurate phylogenetic inference using likelihood analysis on fixed MSAs.
• Current ML software implementations that assume indels as missing data may yield suboptimal results.
• Caution is advised when interpreting phylogenetic trees derived from methods that do not fully account for indel evolution.