Rapid quality evaluation of Meconopsis integrifolia (Maxim.) Franch.: Insights from machine learning techniques and environmental factor analysis

To establish a quality evaluation system for Meconopsis integrifolia based on alkaloid components and to analyze the relationships between active ingredients and environmental factors, 138 samples from 15 origins was evaluated using the TOPSIS-entropy weighting method. Machine learning models (including PLS, SVR and BayesRidge) integrated with near-infrared spectroscopy preprocessing were employed for quality evaluation. Key environmental drivers were then identified through correlation and redundancy analysis. The results from the TOPSIS-entropy weighting indicated that samples originating from the north of Laji Mountain in Qinghai (P4 and P6) demonstrated the highest quality. Machine learning results showed that the optimal model for quality evaluation of different alkaloid components, utilizing both TQ Analyst and Python, achieved high prediction accuracy (RPD>2.8). External validation demonstrated that TQ Analyst exhibited the best predictive performance for OE and NE, whereas the Python-based approach yielded superior results for TA, PE, and AE. Overall, the Python-based models showed enhanced predictive accuracy and robustness, as evidenced by higher prediction rates, reduced MAE and MSE values, and stronger R² scores upon external validation. Spearman correlation and RDA analyses revealed the contents of alkaloid were significantly correlated with altitude, temperature, and precipitation. This study provides a scientific basis for the quality control of M. integrifolia . Additionally, the elucidated relationship between active ingredients and environmental factors lays a theoretical foundation for its protection of excellent germplasm resources and sustainable resource utilization.