Chromatographic Purification of Complex Natural Products as a Decision Problem: Retention Prediction, Adaptive Optimization, and Experimental Feedback.

Chromatographic purification of complex natural products is not only a problem of improving analytical resolution, but also a sequence of decisions involving uncertain targets, variable matrices, limited samples, competing objectives, and scale-dependent constraints. Data-assisted chromatography can support this process, but current evidence does not yet establish the autonomous purification of crude natural-product extracts. This review reframes the field as decision support for staged purification workflows, including analytical profiling, target localization, fraction enrichment, semi-preparative transfer, preparative isolation, and purity or recovery confirmation. We evaluate how retention prediction, global retention modeling, Bayesian optimization, active learning, reinforcement learning, and feedback-driven method development can support different decisions, and where their current boundaries lie. Particular attention is given to chromatographic decision quality, defined by improvements in relevant outcomes such as elution-window selection, target enrichment, fraction purity, recovery, activity retention, solvent burden, robustness, and preparative transferability. The review distinguishes currently implementable tools, near-term human-supervised workflows, and aspirational autonomous systems. Worked examples and evidence-readiness tables are used to show how prediction, optimization, feedback, and expert review can be connected without overstating the maturity of the field.