Leveraging Artificial Intelligence and High-Performance Computing to accelerate biomolecular research and protein design.
This project aims to advance protein design through the development of the IMPRESS framework, while also fostering education and capacity building by utilizing NAIRR resources. It will integrate cutting-edge tools into curricula, exposing non-computing students and researchers to innovative approaches in biological sciences. Co-PI Khare will incorporate parts of the IMPRESS pipeline into an AI-enabled protein science course for graduate students. In partnership with the Research Collaboratory for Structural Biology (RCSB PDB) at Rutgers, the project will offer a crash course on "Introduction to Protein Design using HPC," targeting underrepresented minority students in the RISE program and expanding participation in STEM fields.
Despite advancements in AI and computing, several challenges limit progress:
Determining optimal AI architectures for protein design, efficiently training and fine-tuning models using real-world experimental data, and incorporating diverse data sources such as experiments and simulations. These challenges hinder the accurate and efficient creation of novel protein structures.
IMPRESS increases the impact of AI/ML in protein design by providing advanced systems that enable online coupling of AI and HPC simulations on the NAIRR platform using RADICAL-Cybertools. We focus on real-time evaluation and improvement of models based on experimental data, resulting in high-quality protein structures.
