AutoVap: An Interactive Machine Learning Tool for Predicting the Standard Enthalpy of Vaporization

Abstract

This communication introduces AutoVap (https://autovap.jfcaetano.com), an interactive web application for predicting the standard molar enthalpy of vaporization (Δ_vapH_m°) of chemical structures. Built on an implementation of a supervised machine learning model with RDKit-derived descriptors, AutoVap achieves predictions with a R² test of 96%, across diverse chemical families. The platform allows users to input SMILES notation and obtain both predictions and associated uncertainties through a simple interface. Its validated accuracy and interpretability highlight AutoVap as a reliable and accessible tool, with the potential to advance applications in thermodynamic property estimation, reaction yield prediction, and molecular design.