A38: Machine Learning in Organic Chemistry: Improving the Accuracy of Elimination Reaction Predictions

In recent years, machine learning and artificial intelligence has found unique applications in a plethora of fields and industries. Specifically, in organic chemistry, machine learning algorithms have been used to predict organic synthesis strategies based on previous patented data. Although the current available programs have shown over 90% accuracy on many common industrial reactions, they repeatedly fail on textbook reactions used in organic chemistry. Because of this, the current models available are not useful for student learning purposes. In this project, the main goal is to improve the prediction accuracy for elimination reactions commonly taught in colleges. To achieve an increase in accuracy, the elimination reactions from the Pistachio database were extracted and cleaned to provide a more specialized dataset. In total, two machine learning models were made throughout the duration of this project. The first model was a generic model trained on the whole Pistachio Database which provided a method to compare the new elimination model to the current models available online. After this model was completed, a smaller model was created and tested on a subset of the elimination data extracted. The final model demonstrated a 98% accuracy on the patent testing set of 2,000 reactions and showed promising results when tested manually on specific textbook reactions. Now with the increase in the accuracy of elimination predictions, the new model can be used for educational purposes. It could be added to a program that can be used by students to learn organic synthesis through an online application, or professors could use the tool to create multiple choice problems or demonstrate the effect of solvents on a reaction. Overall, this project demonstrates the versatility of machine learning for use in many different fields, while finding a unique way to improve students’ understanding of organic chemistry.

Author(s): Nikolas Wilson, Chemistry and Chemical Engineering Major

Advisor(s): Benjamin Gung, Department of Chemistry and Biochemistry