Elementary biology textbooks will inform you that all life on Earth is constructed from four categories of particles: proteins, carbs, lipids, and nucleic acids. Furthermore, every set is crucial for each living organism.
However, suppose humankind could essentially demonstrate that these “particles of life,” such as amino acids and DNA bases, may be formed spontaneously in the proper setting? Researchers at the University of Florida are employing the HiPerGator—the most rapid supercomputer in U.S. collegiate education—to test this experiment.
The fastest supercomputer, HiPerGator contains deep neural network models and sizable capacity for graphics processing units, or GPUs (specialized processors designed to expedite graphics renderings), which is revolutionizing the molecular study arena.
Until ten years ago, undertaking research on the evolution and relationships of large groupings of atoms and molecules could only be accomplished utilizing basic computer simulation tests; the computing strength needed to manage the datasets simply wasn’t available.
That is now achievable thanks to HiPerGator. Ph.D. student Jinze Xue from the Roitberg Computational Chemistry Group at UF conducted a comprehensive early Earth chemistry experiment during the winter break in 2023, with the use of over 1,000 A100 GPUs on HiPerGator. This experiment involved performing a molecular dynamics test on 22 million atoms and pinpointing 12 amino acids, three nucleobases, one fatty acid, and two dipeptides. Discovering larger molecules is a significant accomplishment that would not have been feasible in smaller computing systems.
“Our past achievements have enabled us to use Machine Learning and AI to compute energies and forces on molecular systems, generating outcomes that are similar to those of high-level quantum chemistry but around 1 million times quicker,” said Adrian Roitberg, Ph.D., a professor in UF’s Department of Chemistry who has been implementing Machine Learning to examine chemical responses for six years.
“The questions have been posed previously, but due to computational constraints, preceding calculations utilized small numbers of atoms and could not explore the time span necessary to derive results. Now, with HiPerGator, we can do it,” he added.
Erik Deumens, Ph.D., the senior director for UFIT Research Computing, discussed how this full takeover of HiPerGator was viable.
“HiPerGator is uniquely capable of running extraordinarily large ‘hero’ calculations that employ the entire machine, having the potential to lead to breakthroughs in science and scholarship,” said Deumens. “Upon learning about the study conducted by Dr. Roitberg’s group, we proposed running a ‘hero’ test with the code he developed.”
The integration of AI and potent GPUs can facilitate carrying out such data-intensive scientific simulations—calculations that were only a figment of scientists’ imagination a few years ago.
“Using Machine Learning techniques, we devised a simulation employing the entire set of HiPerGator GPUs,” Roitberg mentioned. “We managed to observe the formations of almost every amino acid (alanine, glycine, etc.) and various multifaceted molecules in real-time. It was a highly exhilarating experience.”
This project is part of an ongoing enterprise to unveil how intricate molecules can emerge from fundamental building blocks and to mechanize the process through extensive computer analyses. Roitberg and his research team dedicated numerous hours collaborating with individuals from UFIT. Ying Zhang, UFIT’s AI support manager, was the leader for this experiment.
“Ying assembled a squad comprised of Research Computing staff as well as personnel from NVIDIA to aid our endeavor in scaling compute runs, providing crucial tips and assistance, and expediting the data analysis to the point that the analyses were completed in merely seven hours, instead of the initially expected three days,” Roitberg explained. “We convened weekly, from the preliminary concept through to the final findings, in a very beneficial partnership.”
The trailblazing results and the expeditious delivery by HiPerGator were revolutionary, bringing scientists one step closer to resolving questions about the emergence of intricate molecules. Importantly, the fact that Roitberg managed the computation indicates that
has the ability to reinforce ‘hero runs’ or ‘moonshot calculations’ that advance scientific, engineering, and scholarly endeavors.
“This is an extraordinary opening for UF faculty,” added Roitberg. “Possessing HiPerGator in-house, complemented by the extraordinary personnel willing to go above and beyond to boost researchers in generating revolutionary science as illustrated here, is what makes individuals outside of UF very envious.”
University of Florida
Researchers employ supercomputer to determine if ‘molecules of life’ develop naturally in suitable conditions (2024, February 2)
retrieved 2 February 2024
This document is subject to copyright. Except for equitable dealing for private study or research, no portion may be replicated without the written permission. The content is intended for information purposes only.