COFFEE: a High-Performance Approach to Convex Optimization for Thermodynamic Equilibrium Computations

Understanding complex molecular interactions is a crucial step for advances in synthetic biology, chemical networks, and computational biology. Computing thermodynamic equilibrium in chemical and biological systems provides insight into molecular binding, unbinding, and transformations that occur during reactions, but this problem lacks a closed-form solution and requires computationally intensive numerical methods. Existing tools face challenges with scalability and performance when processing large datasets and lack generality when applied to broader biochemistry problems. To address these issues, we present COFFEE: Convex Optimization For Free Energy Equilibria. Given the input concentrations and free energies of molecules, COFFEE calculates the resulting equilibrium output concentration of the formations. COFFEE is designed to be molecule agnostic and scalable, leveraging a novel application of trust-region optimization methods. COFFEE introduces a high-performance Rust approach and supports both command-line and web-based interfaces. Extensive evaluations against the current state-of-the-art demonstrate an improvement of up to 7 orders of magnitude in constraint adherence and a 2-fold increase in execution speed for large inputs. COFFEE is open source, and a demo is available at https://coffeesolver.dev.