Compiling Stochastic Chemical Trajectories with Elementary Reactions
Can a reaction network’s random behaviour survive when reactions involving many molecules are replaced by elementary steps? This paper gives a constructive answer for finite stochastic mass-action systems supplied with a positive integer population weight and a global upper bound on its drift. The compiler replaces a labelled reaction list of variable dimension and order by a network with at most two molecules on each side of every reaction. Its species, decoder and label map depend only on the reaction list. Every auxiliary carries a nonzero decoded payload, all auxiliary populations start at exactly zero, and directed reachability is preserved for every fine state. Original null reactions and parallel reaction labels remain distinct. One known slowdown and constant rational rates approximate the law of the entire observed trajectory over a requested finite horizon, retaining exact rescaled event-time coordinates, original labels and post-event states. The same output tolerates sufficiently small coordinated, predictable multiplicative errors on every fine channel, even when the errors depend on the complete hidden reaction history. A bounded pending-reaction credit and exact counting identities control population and every actual fine firing throughout the untruncated run, including unsuccessful assembly, release and rare excursions. A separate paired construction supplies exact nominal detailed balance and the same trajectory guarantee for reversible targets with prescribed positive rational activities and a conserved positive integer population weight. This conservative interface is proved separately from the open compiler. The paper distinguishes polynomial binary compiler costs from polynomial numerical physical costs. The latter are charged to an explicit quantity that includes reaction activity at a tail cutoff and can grow rapidly with population and reaction order. These are theorems about ideal mass-action models: the approximation is of path laws, rather than exact simulation, and the results do not establish molecular synthesis, off-graph leakage control, exact decoded equilibrium sampling or practical affordability. An overview of the results, assumptions and proof mechanism is available on the paper’s website page.
Authors
- Harjas Sandhu
Institutions
- The University of Queensland (AU)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-13
- DOI
- https://doi.org/10.5281/zenodo.24170338
- Primary Topic
- Gene Regulatory Network Analysis
- Type
- preprint