DIMPLE: A Tree-Guided Divide-and-Conquer Framework for ILS-Aware Phylogenetic Network Inference on Large-Scale Data

Inferring evolutionary histories of species is crucial for comparative genomics and evolutionary analyses. Reticulate evolution often leads to phylogenetic networks, which better represent relationships than traditional species trees. However, accurately reconstructing large phylogenetic networks is difficult due to the immense number of possible network topologies.

Researchers have introduced DIMPLE, a divide-and-conquer framework for phylogenetic network inference, allowing no restrictions on network level. DIMPLE utilizes a tree-of-blobs (TOB), which symbolizes tree-like relationships between regions where reticulations may exist, while leaving relationships within those regions uncertain.

Within each unresolved region, DIMPLE divides the set of taxa into non-overlapping subsets, each containing a user-defined maximum number of taxa. Several division runs are permitted, and subsets from different runs may share taxa. DIMPLE then infers a subnetwork for each subset, merges them within each region, and combines the resulting networks into a network encompassing the entire taxon set.

Empirical studies using true datasets show that DIMPLE effectively merges networks, providing enhanced reticulation recovery compared to existing methods on higher-level networks. When dealing with estimated datasets, DIMPLE maintains competitive topological accuracy and frequently recovers hybrid ancestry across a range of network complexities.

Its computational demands allow for analyses of hundreds of taxa. Applications to fern and gymnosperm datasets successfully recover documented hybridization and previously acknowledged phylogenetic conflict.