Electronic cigarettes / vaping, are marketed as a safer means to inhale nicotine without the inherent health risks of tobacco. There are currently more than 7,000 flavored vaping e-liquid solutions on the market containing over 200 discrete chemical entities according to the National Academies of Sciences, Engineering, and Medicine in the United State. As temperatures reached in vaping devices can be equivalent to a laboratory pyrolysis apparatus (up to 950 ˚C), the potential for unexpected chemistries to take place on individual components within a vape solution is high. The goal of this proposal is to leverage the power of machine learning with a graphed-convolutional neural network (NN) model to predict the pyrolysis chemistries of known chemicals in vaping solutions, cross correlate the results with experimental mass spectrometry databases, DFT calculated bond dissociation energies (BDEs) and combine with laboratory experiment, toxicity database and globally harmonized system (GHS) classification to generate a vaping health risk predictor. New experimental pyrolysis in-flow methods will be developed to validate the most important predicted results and explore the potential of acid, base and metal catalyzed pyrolysis reactions taking place in the complex vaping mixtures.
Considering such a large number of chemicals involved, it could be expected that many toxic and carcinogenic compounds will be produced (much like tobacco).There is an urgent need for research to study and prevent another tobacco-like health catastrophe as many negative health impacts will not materialize in the vaping population for another decade. With chemistry related machine learning models continuing to advance at speed, their use offers a valuable tool to reveal the longer-term health risks of vaping in advanced of clinical diseases emerging in the general population.