A Collision-based strategy for Network-free Exploration of Complex Molecular Networks

This work presents a stochastic exploration framework for large, implicitly defined chemical reaction spaces that are too large to be generated and stored as explicit molecular networks.

The exploration strategy mimics stochastic chemical kinetics by combining collision-based pair selection with reaction-template instantiation on demand.

In each step, the algorithm first samples molecules to collide, then samples a reaction template, and finally samples a concrete reaction instance among the matches of that template.

This collision-first factorization avoids exhaustive enumeration of all currently possible reactions and enables exploration of large atomistic reaction spaces under open- or closed-system conditions.

We demonstrate the framework on formose chemistry as a case study and analyse both the chemical behaviour reached by the exploration and the computational effects of caching.

The implementation is intended as a general tool for exploratory analysis of generative reaction systems.