Estimating the Q-marker concentrations of Salvia miltiorrhiza via a long short-term memory algorithm using climatic factors and metabolic profiling

Abstract The concentrations of Q-markers are indexes used to evaluate the quality of traditional Chinese medicine. Predicting the yield of Q-markers would aid in decision-making in field management. Climatic factors are significant variables that affect the accumulation of Q-markers. The concentrations of Q-markers are also influenced by other metabolites as precursors or derivatives in the biosynthesis pathway. In this study, a long short-term memory (LSTM) algorithm was used to estimate the concentrations of four Q-markers during the late swelling stage of Salvia miltiorrhiza roots using climatic factors and metabolic profiling. Due to the nonlinearity and heterogeneity of the variables, the correlation analysis for feature selection launched a maximal information coefficient (MIC) analysis to filter the key factors. As a result, the concentrations of the Q-markers varied greatly depending on the active accumulated temperature (AAT), accumulated precipitation (AP) and accumulated sunshine duration (ASD). A strong correlation was found among the tanshinones. The LSTM algorithm made accurate predictions by effectively retrieving the intrinsic characteristics of the climatic factors that fully align with the historic process of metabolic profiling. This approach could help in developing a harvest strategy or taking remedial measures to obtain qualified medicinal materials, especially in extreme weather.