FEEMS

FEEMS infers spatially heterogeneous effective migration surfaces from population genetic data to quantify and visualize gene flow and patterns of isolation-by-distance.

Key Features:

• Gaussian Markov Random Field model: Implements a Gaussian Markov Random Field within a penalized likelihood framework to model spatial genetic variation.
• Per-edge migration parameters: Infers migration parameters for graph edges rather than per-node parameters.
• Penalized likelihood optimization: Uses a penalized likelihood framework enabling efficient optimization of model parameters.
• Computational efficiency: Achieves orders-of-magnitude speedups relative to EEMS while maintaining comparable accuracy.
• Simulation validation: Accurately recovers effective migration surfaces in coalescent simulations, including scenarios with anisotropic gene flow.
• Geographic visualization: Produces effective migration surface representations for mapping spatial patterns of isolation-by-distance.

Scientific Applications:

• Inferring gene flow and isolation-by-distance: Quantifies spatially heterogeneous gene flow and isolation-by-distance from population genetic datasets.
• Analyzing anisotropic migration: Detects and represents anisotropic gene-flow histories in spatial population genetics analyses.
• Empirical population studies: Applied to empirical datasets such as North American gray wolves to interpret spatial population structure.
• Method validation: Used with coalescent simulations to benchmark and validate inference performance.

Methodology:

Implements a Gaussian Markov Random Field within a penalized likelihood framework for efficient optimization, infers migration parameters per graph edge, and is validated using coalescent simulations including anisotropic gene-flow scenarios; performance is compared to EEMS.