Evolutionary distances in the twilight zone--a rational kernel approach.

Area of Science:

• Computational Biology
• Bioinformatics
• Evolutionary Biology

Background:

• Traditional phylogenetic tree reconstruction relies on multiple sequence alignments (MSAs), which degrade in quality with increasing sequence divergence.
• Existing alignment-free methods often lack biological motivation and ignore evolutionary knowledge.
• A key challenge is defining evolutionary distances for divergent sequences that incorporate indel and substitution information without MSAs.

Purpose of the Study:

• To develop a new evolutionary distance metric that overcomes limitations of MSAs and current alignment-free approaches.
• To create a biologically motivated similarity score that models both substitutions and indels.
• To provide an accurate method for phylogenetic reconstruction, especially for highly divergent sequences.

Main Methods:

• Proposed a novel evolutionary distance metric utilizing finite-state transducers.
• Developed a biologically motivated similarity score analogous to pairwise alignments, incorporating substitutions and indels.
• Ensured the similarity score possesses the positive semi-definite property.

Main Results:

• The new evolutionary distance metric demonstrated higher accuracy in phylogenetic reconstruction compared to competing methods in simulation studies.
• Real-world examples confirmed the method's effectiveness, particularly for divergent sequences beyond the 'twilight zone' of alignments.
• The method proved suitable for analyzing large datasets.

Conclusions:

• The proposed finite-state transducer-based evolutionary distance metric offers a significant advancement in phylogenetic tree reconstruction.
• This method provides an accurate and biologically meaningful way to determine evolutionary distances without relying on MSAs.
• The approach is robust for analyzing highly divergent sequences and large-scale genomic data.