Standard maximum likelihood analyses of alignments with gaps can be statistically inconsistent.

Area of Science:

• Evolutionary Biology
• Bioinformatics
• Computational Biology

Background:

• Current phylogenetic methods often treat gaps (insertions/deletions, indels) as missing data.
• The statistical implications of this common practice for phylogenetic inference remain underexplored.

Purpose of the Study:

• To investigate the statistical properties of treating indels as missing data in phylogenetic estimation.
• To determine if standard methods guarantee accurate phylogeny reconstruction under these conditions.

Main Methods:

• Mathematical proof of statistical consistency for maximum likelihood phylogeny estimation.
• Analysis under a general model of sequence evolution, assuming a true alignment.

Main Results:

• Maximum likelihood phylogeny estimation is statistically inconsistent when indels are treated as missing data.
• Accurate phylogenetic inference is not guaranteed, even with perfect alignments and extensive sequence data.

Conclusions:

• Standard phylogenetic estimation techniques have potentially unfavorable statistical properties when handling indels as missing data.
• Developing methods that explicitly model indel events is crucial for improving phylogenetic accuracy.