A machine learning approach to study plant functional trait divergence.

Area of Science:

• Plant Ecology
• Evolutionary Biology
• Bioinformatics

Background:

• Plant functional traits are crucial for understanding ecological strategies and species differentiation.
• Multivariate trait spaces are complex, making it challenging to pinpoint key drivers of phenotypic divergence.
• The genus *Helianthus* (sunflowers) provides a model system with multiple radiations for studying evolutionary patterns.

Purpose of the Study:

• To develop and apply a machine learning (ML) approach for identifying traits most important for interspecific phenotypic divergence.
• To investigate trait importance at both the genus and clade levels within *Helianthus*.
• To assess the repeatability of evolutionary divergence by comparing trait importance across independent radiations.

Main Methods:

• Applied descriptive and predictive ML models, including random forest and gradient boosting machine classifiers.
• Utilized recursive feature elimination to identify significant traits.
• Analyzed trait data for the genus *Helianthus* at the genus and three major clade levels.

Main Results:

• ML models achieved high accuracy in predicting species identity from functional traits.
• Significant differences in the importance of functional traits were found between the genus and clade levels.
• These differences indicate distinct major axes of phenotypic divergence at different taxonomic levels.

Conclusions:

• ML approaches effectively identify key divergent traits, offering insights into evolutionary predictability and repeatability.
• Comparing trait divergence across parallel radiations within *Helianthus* illuminates evolutionary processes.
• This methodology is broadly applicable to plant science, from basic research on interspecific divergence to applied studies of intraspecific variation.