This project aims to enhance molecular structure prediction by integrating AI-driven methods with High-Performance Computing (HPC) and spectroscopic data. Accurate structure prediction is essential for drug discovery, materials science, and energy solutions, enabling the development of safer therapies, sustainable materials, and improved energy storage technologies.
Challenges and AI Solutions
Identifying molecular structures through traditional methods is time-consuming due to the vast and complex "chemical space" of possible structures. AI-driven models offer a solution by efficiently learning patterns from experimental data, significantly accelerating structure prediction.
Machine Learning and HPC Integration
By leveraging spectroscopic data such as infrared (IR) measurements, Machine Learning (ML) techniques allow direct molecular structure prediction. HPC enhances these models by integrating quantum mechanics and large-scale datasets, accelerating both structure prediction and optimization.
Research Focus
The primary focus of our project is to develop a structure prediction method for molecules based on computed (e.g., IR) spectra. While techniques like IR generate detailed molecular data, analyzing this information to determine structures remains complex. AI models trained on datasets of IR measurements can identify molecular structures with high accuracy and efficiency. Our ultimate goal is to develop an ML-based model trained on highly accurate computational quantum chemistry data, capable of predicting molecular structures directly from IR spectra.
Impact and Applications
Beyond structure prediction, we are developing universal, open-source software tools adaptable to various spectroscopic data types. These tools are designed to make chemical predictions more accessible by leveraging ML and HPC, benefiting a wide range of scientific fields, including chemistry, materials science, and pharmaceuticals. By combining AI, HPC, and quantum chemistry, this interdisciplinary approach advances molecular behavioral understanding across fields.
