基于辛普森积分法的红外光谱分析冬虫夏草中腺苷含量OA

To solve the limitations of traditional adenosine content determination methods,such as complex sample preparation and time-consuming operations,in this study,Fourier transform infrared(FTIR)spectroscopy combined with chemometric methods was employed to establish a rapid detection model for adenosine content in Cordyceps sinensis.A rapid,efficient,and non-destructive quantitative method was proposed based on FTIR spectroscopy integrated with Simpson's integration rule.Spectral data were collected from 77 samples of C.sinensis.Through statistical analysis,1 365 significant data points were screened from 1 868 original data points,determining five critical wavenumber regions(399.19-833.29,1 033.36-1 380.46,1 710.13-1 790.89,1 870.56-3 020.78,and 3 980.46-3 999.71 cm-1).Four types of spectral characteristic variables—peak area(A),full width at half maximum(FWHM),area-to-height(H)ratio(A/H),and mean absorbance of the spectral band(MBA).A total of 62 samples(80%)were allocated as the training-validation set for 5-fold cross-validation to establish Multiple Linear Regression(MLR)and Partial Least Squares Regression(PLSR)models;the remaining 15 samples(20%)served as an independent test set for model generalization validation.FTIR characteristic analysis revealed that the principal absorption peaks of C.sinensis were located at 3 289.17 cm-1(O—H stretching),2 928.53 cm-1(asymmetric C—H stretching),and 1 652.72 cm-1(amide I band).Modeling results demonstrated that the A/H ratio feature exhibited optimal performance on the calibration set(MLR:R²=0.941±0.029),but suffered from severe overfitting(ΔR²=0.225);the PLSR-MBA combined model achieved the best cross-validation performance(R²=0.807±0.031,RMSE=0.189±0.005)with minimal overfitting(ΔR²=0.078).Independent test set validation further confirmed the superior predictive accuracy and robustness of the PLSR-MBA model(R²=0.886,RMSE=0.107),significantly outperforming the MLR-A/H model(R²=0.845,RMSE=0.169).The PLSR-MBA combined model achieves an optimal balance among fitting accuracy,generalization capability,and overfitting resistance,and is recommended as the preferred method for rapid quantitative analysis of adenosine content in C.sinensis.This study provides theoretical basis and practical guidance for feature extraction and modeling strategy selection in spectral quantitative analysis of traditional Chinese medicines.