simlandr: Simulation-Based Landscape Construction for Dynamical Systems

We present the simlandr package for R, which provides a set of tools for constructing potential landscapes for dynamical systems using Monte Carlo simulation. Potential landscapes can be used to quantify the stability of system states. While the canonical form of a potential function is defined for gradient systems, generalized potential functions can also be defined for non-gradient dynamical systems. Our method is based on the potential landscape definition from the steady-state distribution, and can be used for a large variety of models. To facilitate simulation and computation, we introduce several novel features, including data structures optimized for batch simulations under varying conditions, an out-of-memory computation tool with integrated hash-based file-saving systems, and an algorithm for efficiently searching the minimum energy path. Using a multistable cell differentiation model as an example, we illustrate how simlandr can be used for model simulation, landscape construction, and barrier height calculation. The simlandr package is available at https://CRAN.R-project.org/package=simlandr, under GPL-3 license.