LC–MS/MS Software for Screening Unknown Erectile Dysfunction Drugs and Analogues: Artificial Neural Network Classification, Peak-Count Scoring, Simple Similarity Search, and Hybrid Similarity Search Algorithms

Screening and identifying unknown erectile dysfunction (ED) drugs and analogues, which are often illicitly added to health supplements, is a challenging analytical task. The analytical technique most commonly used for this purpose, liquid chromatography–tandem mass spectrometry (LC–MS/MS), is based on the strategy of searching the LC–MS/MS spectra of target compounds against database spectra. However, such a strategy cannot be applied to unknown ED drugs and analogues. To overcome this dilemma, we have constructed a standalone software named AI-SIDA (artificial intelligence screener of illicit drugs and analogues). AI-SIDA consists of three layers: LC-MS/MS viewer, AI classifier, and Identifier. In the second AI classifier layer, an artificial neural network (ANN) classification model, which was constructed by training 149 LC–MS/MS spectra (including 27 sildenafil-type, 6 vardenafil-type, 11 tadalafil-type ED drugs/analogues and other 105 compounds), is included to classify the LC–MS/MS spectra of the query compound into four categories: i.e., sildenafil, vardenafil, and tadalafil families and non-ED compounds. This ANN model was found to show 100% classification accuracy for the 187 LC–MS/MS modeling and test data sets. In the third Identifier layer, three search algorithms (pick-count scoring, simple similarity search, and hybrid similarity search) are implemented. In particular, the hybrid similarity search was found to be very powerful in identifying unknown ED drugs/analogues with a single modification from the library ED drugs/analogues. Altogether, the AI-SIDA software provides a very useful and powerful platform for screening unknown ED drugs and analogues.