Identification of Phosphodiesterase type 5 inhibitors (PDE5is) analogues using surface-enhanced Raman scattering and machine learning algorithm

Phosphodiesterase type 5 inhibitors (PDE5is), primarily used for the treatment of erectile dysfunction, have been severely misused in recent years, posing a threat to public health and safety. This study developed a method that combines Surface-enhanced Raman spectroscopy (SERS) with machine learning algorithms to rapidly identify different PDE5is types. A total of 948 SERS spectra from 79 PDE5is were collected using gold nanoparticles (AuNPs) as the enhancement agent, and dimensionality reduction was performed through principal component analysis (PCA). Subsequently, six traditional machine learning models, partial least squares discriminant analysis (PLS-DA), orthogonal partial least squares discriminant analysis (OPLS-DA), support vector machines (SVM), k-nearest neighbors (KNN), random forest (RF), and multilayer perceptron (MLP) were applied for data classification and identification. Results showed that the MLP model achieved the highest classification accuracy of 99.65 %, with only 1.82 % of the samples misclassified from thiosildenafil to sildenafil analogues, significantly outperforming the other models. This method offers a rapid, cost-effective, and accurate alternative for the detection of PDE5is in health foods, with implications for improving regulatory oversight and public health safety.