Neural Network Potentials for Low-Energy 3D Structure Generation and Reactivity Prediction

Neural Network Potentials for Low-Energy 3D Structure Generation and Reactivity Prediction

Explore a comprehensive 50-minute talk on neural network potentials for low-energy 3D structure generation and reactivity prediction. Delve into the challenges of obtaining optimal molecular structures and learn about the Python-based Auto3D package, which automates isomer enumeration, 3D building, geometry optimization, and ranking processes. Discover the ANI-2xt model, an extension of ANI trained on tautomer-rich datasets, and its improved performance in tautomeric reaction energy calculations. Gain insights into the background, uncertainty in stereochemistry, neural network potentials, and Auto3D's overview. Examine a practical example of Gibbs free energy calculations for tautomerization reactions and conclude with an Auto3D tutorial.