Speaker
Description
Neutral birth–death–mutation dynamics in finite populations can produce patterns in low-dimensional summaries of trait-distance space (e.g., mean pairwise distance) often interpreted as signatures of selection, making inference of selection from snapshot data fundamentally ambiguous. We introduce a distance-space framework for finite-length, one-parent Moran populations that separates neutral dynamics from selective bias directly at the level of trait distances.
Our key observable is a parent-bias signal defined as the difference between two distance repertoires: distances between reproducing parents (with higher fittness) and individuals randomly selected for death, and distances obtained by uniformly sampling birth–death pairs from the population. The resulting ODE for pairwise-distance dynamics decomposes into an exact neutral baseline and a measurable parent-bias forcing term. The baseline is determined by mutation transition probabilities at birth, while the forcing term is obtained by applying the same mutation kernel to the parent-bias signal.
By using forcing inputs obtained from agent-based simulations with recorded parental histories, the ODE reproduces both selection-driven convergence and mutation-driven relaxation of pairwise distances. A first-moment summary of the parent-bias signal provides a quantitative readout of selection timing and strength. The framework applies whenever parent–offspring relationships (or equivalent parental assignments) are available.
