Statistical Inconsistency of Maximum Parsimony for k-Tuple-Site Data

Area of Science:

• Phylogenetics and evolutionary biology
• Computational biology and bioinformatics

Background:

• Phylogenetics aims to reconstruct the evolutionary history of life, often visualized as the "Tree of Life."
• Maximum parsimony is a criterion used for inferring evolutionary relationships from DNA sequences.
• Existing methods can produce inaccurate estimates, with long branch attraction being a known issue leading to statistical inconsistency.

Purpose of the Study:

• To investigate the statistical consistency of k-tuple analyses in phylogenetic tree reconstruction.
• To determine if k-tuple analyses offer an improvement over site-based analyses for more than four taxa.
• To analyze the behavior of maximum parsimony with k-tuple data for five taxa.

Main Methods:

• Utilized k-tuple (blockwise) analyses of DNA sequence data.
• Focused on phylogenetic tree reconstruction using maximum parsimony.
• Examined cases with five taxa, using 2- and 3-tuple data with two and four-element alphabets.

Main Results:

• The equivalence between single-site and k-tuple analyses, observed in four-taxon cases, does not hold for five taxa.
• Maximum parsimony demonstrates statistical inconsistency when applied to k-tuple site data with five taxa.
• The use of k-tuples does not resolve the statistical inconsistency issues of maximum parsimony for larger taxon sets.

Conclusions:

• K-tuple analyses do not inherently resolve the statistical inconsistency problems of maximum parsimony in phylogenetics.
• The findings highlight limitations of maximum parsimony and k-tuple methods for reconstructing complex evolutionary histories with more than four taxa.