Population genetic simulation: Benchmarking frameworks for non-standard models of natural selection

Area of Science:

• Population genetics
• Computational evolutionary biology
• Bioinformatics

Background:

• Population genetic simulations are crucial for studying complex evolutionary and demographic models.
• Advancements in tree sequence recording enable merging coalescent and forward simulation efficiencies.
• No comprehensive benchmarking exists for these advanced simulation frameworks.

Purpose of the Study:

• To evaluate and compare different simulation workflows for population genetics.
• To identify the most resource-efficient simulation framework for complex evolutionary scenarios.
• To benchmark coalescent and forward simulation methods with tree sequence recording.

Main Methods:

• Utilized msprime (coalescent) and SLiM (forward) simulators.
• Evaluated three key simulation stages: burn-in, forward simulation with selection, and statistics computation.
• Assessed memory usage and computation time for each workflow.

Main Results:

• The fastest framework combines coalescent and forward simulation with tree sequence recording.
• This hybrid approach is over twenty times faster than classical forward simulations.
• The optimal framework requires six times more memory compared to traditional methods.

Conclusions:

• Efficient simulation workflows, particularly those integrating coalescent and forward methods with tree sequence recording, substantially improve modeling of complex evolutionary scenarios.
• The choice of simulation framework depends on balancing speed and memory resource availability.
• This research provides essential benchmarks for selecting appropriate computational tools in population genetics.