CorleoneGame: A library of biological dynamic games with a companion Julia solver

Abstract

Biological agents, from regulatory modules and microbial strains to organisms and populations, act on a common environment while each responds to its own costs, benefits, and limits. A dynamic game makes such interactions explicit by assigning who controls which decision, what each participant values, and which restrictions constrain unilateral deviations. We present a library of 25 biological dynamic games with 2 to 10 players spanning molecular, cellular, organismal, and population scales and competitive, antagonistic, mutualistic, and regulatory interactions. Instances cover single seasons, cycles that return to their initial state, or episodic tasks; seven include duration choice and some players' feasible sets are coupled. For each game we provide biological motivation, an executable Julia specification with player-owned controls, objectives and restrictions, a parameter instance, and a reference open-loop generalized Nash equilibrium candidate with interpreted trajectories. The solver CorleoneGame.jl computes candidates by sequential optimal-control best responses and audits them for feasibility and profitable unilateral deviations on a coarse and a refined control grid. All 25 reference computations satisfy the audit tolerances on both grids. Using a plant--herbivore game as a running example, we show how an executable specification yields an audited reference candidate that qualitatively reproduces a published prediction, how biological hypotheses become instance variants with different candidates, and how the candidates generate synthetic data for inverse games.