The fields of ecology and evolutionary biology are fundamentally structured by relationships represented as trees and graphs. Phylogenetic trees delineate evolutionary ancestry, while ecological interaction networks and gene-flow graphs capture intricate relationships such as predation, pollination, dispersal, and hybridization. However, real-world datasets present formidable challenges: incomplete sampling, noisy measurements, and co-occurring processes that entangle evolutionary, ecological, and environmental signals across different scales. The advent of large, multimodal datasets, integrating genomics, traits, climate variables, microbiomes, and remote sensing, adds complexity that often exceeds the capacity of traditional models.

Recent advancements in artificial intelligence, notably deep learning techniques designed for structured data, have introduced transformative opportunities. Graph Neural Networks (GNNs), tree-structured transformers, and geometric deep learning approaches excel at encoding complex topologies, integrating heterogeneous data types, and uncovering hidden patterns within these networks. Applications now span prediction of missing links or traits, inference of ancestral states, exploration of latent eco-evolutionary regimes, and simulation-based approaches for scalable phylogenetic or network analysis. Despite this promise, hurdles such as model interpretability, rigorous uncertainty quantification, cross-system transferability, and mechanistic interpretability persist. A deeper synthesis between machine learning innovations and eco-evolutionary theory is critical for meaningful, actionable insights.

This Research Topic aims to catalyze the development and critical evaluation of learning methods that treat trees and graphs as essential, interpretable objects in ecology and evolution. It seeks contributions that bridge deep learning with biological theory, enhance causal and uncertainty understanding, and provide robust tools for predicting, inferring, and interpreting real-world biological complexity.

To gather further insights in the intersection of machine learning, phylogenetics, and eco-evolutionary systems, we welcome submissions focusing on advances and applications that address, but are not limited to, the following themes:
- GNNs for ecological interaction networks (e.g., food webs, pollination, microbiomes)
- Deep learning for phylogenetic comparative methods (traits, rates, correlated evolution)
- Differentiable phylogenetics and likelihood-free inference (simulation-based, amortized approaches, SBI)
- Phylogeography and landscape graphs (dispersal, connectivity, resistance surfaces)
- Graph representation learning for community assembly and metacommunities
- Multimodal data integration on trees/graphs (e.g., genomics, traits, climate, remote sensing)
- Methods for handling uncertainty, calibration, and robustness to missing or biased data
- Explainability and causal discovery in eco-evolutionary graph models

Submissions reflecting these themes are encouraged in all suitable formats, including Original Research, Methods, Reviews, Mini Reviews, Hypothesis & Theory, Perspectives, Technology Reports, Data Reports, and Brief Research Reports. We particularly welcome reproducible pipelines, open-source benchmarks, and innovative approaches that enhance transparency and interpretability.

Are you interested in contributing a manuscript to this Research Topic?

Learning on Trees and Graphs: Deep Learning & Graph Neural Networks for Phylogenetics and Eco‑Evolutionary Systems