Delineating reef fish trophic guilds with global gut content data synthesis and phylogeny

Area of Science:

• Ecology
• Marine Biology
• Computational Biology

Background:

• Accurate assessment of species' trophic niche is crucial for understanding food web dynamics.
• Expert opinion on trophic guilds is often inconsistent, especially in hyperdiverse ecosystems like coral reefs.
• Disagreement among experts on trophic guild assignment for over 20% of species hinders cross-study comparability.

Purpose of the Study:

• To develop a quantitative, unbiased, and reproducible method for defining trophic guilds.
• To utilize machine learning to predict probabilities of trophic interactions with high accuracy.
• To provide a framework for advancing reproducibility in trait-based ecology.

Main Methods:

• Synthesized global gut content data from 13,961 tropical coral reef fishes (615 species).
• Employed network analysis to identify 8 distinct trophic guilds.
• Used Bayesian phylogenetic modeling to predict trophic guilds based on phylogeny and body size.
• Applied machine learning to predict pairwise trophic interactions.

Main Results:

• Developed a reproducible trophic categorization scheme.
• Machine learning models achieved a misclassification error of less than 5% for predicting trophic interactions.
• Demonstrated that trophic guilds can be predicted using phylogeny and maximum body size.
• Generated high-resolution probabilities for pairwise trophic interactions.

Conclusions:

• The proposed framework offers a viable approach to accurately define trophic guilds and interactions in complex ecosystems.
• This method enhances reproducibility in trait-based ecology, applicable beyond coral reef fishes.
• Provides a robust tool for understanding predator-prey dynamics in highly diverse environments.