Imputing missing distances in molecular phylogenetics.

Area of Science:

• Molecular Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Missing data is a common challenge in molecular phylogenetics, hindering accurate phylogenetic reconstruction.
• Existing distance imputation methods often lack accuracy or rely on restrictive assumptions like molecular clocks.

Purpose of the Study:

• To develop an accurate distance imputation method for distance-based phylogenetic reconstruction.
• To address the limitations of current methods in handling missing data in phylogenetic analyses.

Main Methods:

• A novel least-square method combined with multivariate optimization was developed.
• The method imputes multiple missing distances in distance matrices or from aligned sequences with missing genes.
• It handles cases where sequences share no homologous sites.

Main Results:

• Phylogenetic trees can be reliably inferred from distance matrices with approximately 10% missing distances.
• The accuracy of trees reconstructed with imputed data is comparable to those from complete datasets.
• The new method outperforms a previous approach by not assuming a molecular clock and showing comparable accuracy to maximum likelihood methods.

Conclusions:

• The developed distance imputation method provides an accurate and flexible solution for molecular phylogenetics.
• It enables robust phylogenetic tree inference even with substantial missing data.
• The method is implemented in the freely available DAMBE software.