Inferring Phenotypic Trait Evolution on Large Trees With Many Incomplete Measurements.

Area of Science:

• Phylogenetic Comparative Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Comparative biologists analyze trait covariation across related taxa, requiring control for evolutionary history to prevent spurious inferences.
• Challenges include increasing difficulty in obtaining complete datasets with more taxa, necessitating data imputation or integration.
• Existing methods for controlling evolutionary history scale poorly with increasing numbers of taxa.

Purpose of the Study:

• To develop a scalable inference technique for analyzing trait covariation in the presence of missing data and evolutionary history.
• To extend the multivariate Brownian diffusion (MBD) model to incorporate sampling error or residual variance.
• To provide a computationally efficient solution for phylogenetic comparative methods and matrix-normal distribution likelihood calculations.

Main Methods:

• Proposed an inference technique that analytically integrates out missing measurements.
• Utilized a post-order traversal algorithm under a multivariate Brownian diffusion (MBD) model for linear scaling with taxa number.
• Extended the MBD model to account for sampling error and non-heritable residual variance.

Main Results:

• Achieved computational efficiency increases of up to two orders of magnitude compared to current best practices.
• Demonstrated the method's applicability across diverse datasets including mammalian life history, prokaryotic traits, and HIV infection data.
• The approach generalizes to computing likelihoods for matrix-normal and multivariate normal distributions with missing data.

Conclusions:

• The proposed method offers a computationally efficient and scalable solution for phylogenetic comparative analyses with missing data.
• The technique effectively controls for shared evolutionary history while handling data limitations.
• This approach has broad applicability in evolutionary biology and statistical modeling of complex datasets.