Category: Pop. Genetics

Summary: Finding how much cross-module reactivation leak is needed to erase seed-bank-protected localization in a GPU-accelerated modular quasispecies model.

Dormancy can protect modular genetic structure by storing lineages in a reservoir that buffers ongoing mutation and recombination. This experiment asks when that protection breaks down by measuring how much cross-module reactivation leak is required before the dominant quasispecies mode loses the localization maintained by the seed bank.

The model builds a dense symmetric operator over active and dormant states, then uses GPU-accelerated eigensolves, iterative deepening, and bisection on the leak parameter. The threshold is defined through loss of modular localization rather than only through overall growth, so the result focuses on the structural failure of seed-bank buffering.

That combines modular fitness landscapes, recombination, and dormancy in a single finite-size threshold problem. The aim is to quantify how robust dormant-state protection really is once reactivation no longer cleanly respects modular boundaries.

Method: GPU dense symmetric eigensolve with iterative deepening and bisection on cross-module reactivation leak in a coupled active-dormant quasispecies operator.

What is measured: Critical leak threshold, localization of the leading mode, dormant-buffer robustness, and bracket width.