Network free models

[isaacsas]

https://discourse.julialang.org/t/using-catalyst-jl-with-large-reaction-networks/123061/2 is, I think, asking about something I've thought we need for a while -- a way to specify network free models (and corresponding Gillespie solvers in JumpProcesses). At the Catalyst level we'd need to decide how to represent such models, perhaps pulling from Bionetgen's modeling language or some of the dedicated sys bio language (such as the Kappa language). It might be good to use this thread to collect resources for thinking about this.

[isaacsas]

I'd think the main feature a network free solver needs is the ability to have a dynamic dependency graph / set of reactions, which seems suggestive that table-based methods might work well there (since they have fast updating / removal of reactions). But I haven't played with these methods before, so perhaps there are better approaches in the literature.