This thesis presents a detailed framework and implementation (the Multistrand simulator) for simulating the stochastic kinetics of multi-strand nucleic acid systems.

...the core innovation lies in developing data structures (loop graph, move tree) and algorithms for efficiently handling a continuous-time Markov process on a state space characterized by dynamic topology... where transitions (moves) are generated by local changes... within these structures.

With modern computational power and the rise of Graph Neural Networks (GNNs), one could envision training GNNs to predict move propensities or even generate valid moves directly from the loop graph representation, potentially replacing the more computationally expensive explicit enumeration and energy calculations described in the thesis (O(N^2) move generation in the worst case).

Unlike typical simulations that model interactions between fixed entities or on static graphs, this thesis provides a concrete, low-level mechanism for handling the combinatorial explosion and computational challenges introduced by the creation and destruction of bonds between entities, leading to a continuously evolving system graph.