Intelligent identification technology of zidovudine and its isomers based on active low-frequency terahertz...

Zidovudine, adenosine, and vidarabine are three isomeric compounds with highly similar physicochemical properties, making them prone to confusion during pharmaceutical manufacturing, storage, and transportation. Therefore, the development of identification methods with high specificity and sensitivity for such isomeric compounds is of great significance for ensuring drug quality and clinical safety. However, conventional detection methods suffer from limitations such as high cost, long processing time, or insufficient resolution, rendering drug identification a challenging issue. This study introduces a non-destructive and intelligent detection method by integrating terahertz time-domain spectroscopy (THz-TDS) with machine learning techniques. Transmission spectra of zidovudine, adenosine, and vidarabine were collected using a THz-TDS system and converted into absorbance spectra. Key features were extracted via t-distributed stochastic neighbor embedding (t-SNE), isometric mapping (Isomap), and principal component analysis (PCA). These were used to train support vector machine (SVM) and k-nearest neighbors (KNN) classifiers. Among the obtained models, the PCA-KNN combination achieved the most outstanding performance, with a classification accuracy of 100%, realizing precise discrimination of the three isomers. This method not only meets the quality control requirements of pharmaceutical standards and industrial production but also serves as a valuable reference for THz spectral detection of related drugs. It demonstrates promising potential for applications in precise drug identification and recognition.