RapidKnock enables fast linear programming-based knockout search for microbial strain design

Identifying reaction knockout strategies that redirect metabolic flux toward overproduction of a desired metabolite is a central computational challenge in metabolic engineering. Classical strain design tools, e.g. OptKnock, rely on mixed-integer linear programming, are typically run with commercial solvers for genome-scale problems, and return a single solution. Exhaustive enumeration methods such as FastKnock avoid the commercial solver dependency by relying on linear programming alone, but at the cost of execution times that scale into the hours-to-days range. Here we describe RapidKnock, an iterative search algorithm that relies on linear programming, using flux variability analysis to prune the knockout candidate space and flux balance analysis to identify and lock in high-yield knockout combinations. On the Escherichia coli iAF1260 genome-scale metabolic model under succinate-production conditions, RapidKnock identified knockout strategies faster than OptKnock, by at least 1.45-fold in the larger quadruple-knockout case, while returning a ranked pool of alternative solutions. On the Aspergillus niger iJB1325 genome-scale metabolic model, used as a computational stress test for oxaloacetate overproduction, RapidKnock identified viable knockout strategies where both OptKnock and FastKnock failed to return any solution within the allotted computation time. RapidKnock runs on free open-source solvers and, on the cases tested here, returns results faster and with more ranked alternatives than OptKnock and FastKnock.