Automating Observable Closure Analysis for Graph Transformation Models with MolFoundry

Abstract

Graph transformations model changing systems, with applications in software, biology and chemistry, among others. We introduce MolFoundry, a graph transformation framework, and present a workflow for observable analysis that tracks the dynamics of local patterns through rule composition. Analysing rule compositions avoids the need to enumerate a potentially infinite state space. Building on rule-algebraic stochastic rewriting, MolFoundry's kernel composes rules with nested application conditions, including missing-edge constraints. Given rules, rates and query patterns, the algorithms iteratively identify auxiliary pattern counts and search for closure. Using exact counting identities supplied by graph-family backends, successful searches yield closed ordinary differential equations for average counts or, with justified individual jumps, finite event models describing joint changes and rates. We evaluate conditional composition and automated discovery of auxiliary observables against published calculations and independent rule application. Applications to aldol addition, copolymerization and chlorination show how the framework yields quantitative predictions directly from molecular graph rules.