Parameter Scaling in Population Genetics Simulations may Introduce Unintended Background Selection: Considerations...

Area of Science:

• Population genetics
• Computational biology
• Evolutionary genetics

Background:

• Scaling is widely used in population genetic simulations to enhance computational efficiency.
• The impact of scaling on diversity estimates and comparability with empirical data remains under-examined.

Purpose of the Study:

• To systematically investigate the effects of scaling on genetic diversity estimates and simulation accuracy.
• To compare scaled simulations with unscaled models and empirical data across species with different demographic parameters.

Main Methods:

• Simulations were conducted for modern humans and Drosophila melanogaster, species with distinct population sizes and generation times.
• Analyzed the influence of scaling factors on coalescence, runtime, memory usage, diversity estimates, site frequency spectra, and linkage disequilibrium.
• Evaluated the impact of simulated segment length and burn-in duration on simulation metrics.

Main Results:

• Scaling significantly improves computational efficiency but can distort genetic diversity and variant dynamics at large scaling factors.
• Strongly scaled simulations may exhibit amplified background selection and purging of linked mutations, deviating from intended models.
• A common heuristic burn-in length (10N generations) proved insufficient for full coalescence and altered linkage disequilibrium patterns.

Conclusions:

• While scaling enhances computational performance, excessive scaling can lead to inaccurate genetic diversity and evolutionary dynamics.
• Burn-in length and scaling factors require careful consideration to ensure simulation fidelity.
• Bespoke scaling strategies are recommended for nonmodel species simulations to mitigate potential distortions and ensure comparability with empirical data.