A Two-Stage Screening-to-Optimization Approach with Mechanistic Model Analysis: Enhancing Anthocyanin in Lettuce Without Yield Loss

Enhancing anthocyanin accumulation in red-leaf lettuce grown in plant factories often incurs yield penalties. Here we propose a two-stage screening-to-optimization framework integrated with mechanistic modeling to resolve this tradeoff. In Stage 1, comparative experiments confirmed that UV-A is more compatible with growth and pigmentation than UV-B, and identified 'Lollo Rosso' as a highly responsive cultivar. In Stage 2, optimization experiments showed that L6D6 (6 h day^-1 for 6 days) increased the total anthocyanin per plant by 19.9% while maintaining fresh weight. Motivated by observed nonlinear phenomena including biomass overcompensation, circadian disruption under night irradiation, and ontogeny-dependent vulnerability, we developed a six-state ordinary differential equation (ODE) model that integrates reactive oxygen species (ROS) dynamics with stress damage-repair processes. A key innovation is the explicit representation of carbon competition between growth and antioxidant defense, where AOX synthesis consumes carbon from the buffer pool, creating a physiologically meaningful growth-defense tradeoff supported by the Growth-Differentiation Balance Hypothesis. The model achieved high accuracy in an independent validation set that included extreme doses (errors ≤ 10.6%, with 11 of 12 metrics < 10%), supporting the physiological necessity of the introduced mechanisms. Global optimization based on the calibrated model predicted that 9 h day^-1 for 4 days is the theoretical optimum, potentially increasing total anthocyanin by 38.3% with minimal fresh-weight reduction (-2.4%), substantially outperforming the best experimental treatment. This quantitative mechanistic framework provides a scientific basis for designing precise stress-light recipes in controlled-environment agriculture.