Machine learning based imputation techniques for estimating phylogenetic trees from incomplete distance matrices.

Area of Science:

• Evolutionary Biology
• Genomics
• Bioinformatics

Background:

• Species tree inference is crucial in evolutionary biology, but large genomic datasets present challenges.
• Existing distance-based phylogenetic methods like Neighbor Joining (NJ) and UPGMA require complete distance matrices, hindering analysis of incomplete data.
• Handling missing data is a key challenge in leveraging large-scale genomic information for evolutionary studies.

Purpose of the Study:

• To develop and evaluate novel, accurate methods for imputing missing data in distance matrices for phylogenetic tree construction.
• To address the limitations of existing methods in handling incomplete genomic datasets.
• To advance the field of phylogenetic inference by improving the handling of missing data.

Main Methods:

• Introduced two machine learning-based distance imputation techniques: matrix factorization and autoencoder-based deep learning.
• Evaluated the performance of these methods on both simulated and real biological datasets.
• Assessed the scalability of the methods for large datasets with numerous taxa.

Main Results:

• The proposed machine learning methods demonstrate high accuracy in distance imputation.
• These techniques match or surpass the performance of existing alternative imputation methods.
• The methods are scalable to large datasets and effectively handle substantial amounts of missing data.

Conclusions:

• This study demonstrates the efficacy of deep learning for distance matrix imputation in phylogenetics.
• The findings advance the state-of-the-art in phylogenetic tree construction, particularly in the presence of missing data.
• The developed imputation methods are publicly available as open-source software.