Mechanistic multi-objective optimization of ion-pair reversed-phase liquid chromatography for oligonucleotide...

Oligonucleotides (ONs) play a vital role in diagnostics and therapeutics, with ion-pair reversed-phase liquid chromatography (IP-RPLC) being a key method for their purification. This study presents a numerical approach to optimize ON purification using both single-objective (productivity) and multi-objective (productivity and yield) optimization. A transport-dispersive column model incorporating a gradient-modified Langmuir kinetic adsorption-desorption isotherm was employed to describe the separation dynamics. The model was validated against experimental elution profiles of a 20-mer ON and its six most abundant shortmer impurities, and then used to optimize injection volume and gradient slope under varying purity constraints. A hybrid optimization strategy combining simulated annealing (stochastic) with the simplex algorithm (deterministic) was applied. The results showed that productivity was maximized at the highest injection volume, while higher purity constraints required shallower gradient slopes. Multi-objective optimization yielded Pareto fronts with unexpected shapes, including inflection points or local minima. We hypothesize that slow adsorption-desorption kinetics limited the model's ability to accurately describe separations across a broad range of gradient slopes. To test this, two new models were calibrated using only shallow or steep gradient data. The resulting Pareto fronts displayed no inflection points or local minima, supporting our hypothesis.