AI Discovers That Several Life Foundations, Including DNA, Can Naturally Emerge

Published by Adrien - Thursday, February 15, 2024 - Other Languages: FR, DE, ES, PT
Source: University of Florida

The University of Florida announces a major breakthrough in understanding the formation of life's molecules. Through innovative experimentation, researchers have used the supercomputer HiPerGator to demonstrate that essential life molecules, such as amino acids and DNA bases, can naturally form under specific conditions.


The HiPerGator supercomputer, recognized as the fastest in the American academic milieu, has enabled a new step in molecular research thanks to its artificial intelligence models and its exceptional capability in graphics processing units (GPU). These tools have made it possible to study the interactions and evolution of vast sets of atoms and molecules, a task previously unthinkable with available computing capacities.

Jinze Xue, a Ph.D. student at the University of Florida, conducted a prebiotic chemistry experiment during the winter holidays of 2023. Using more than 1000 A100 GPUs, the experiment managed to identify 12 amino acids, three nucleobases, one fatty acid, and two dipeptides among 22 million atoms. This discovery marks a significant progress, revealing the formation of complex molecules that would not have been detectable with less powerful computing systems.

The success of this research relies on the use of machine learning and artificial intelligence to calculate energies and forces acting on molecular systems. According to Adrian Roitberg, a professor in the Department of Chemistry at the University of Florida, these methods provide results comparable to those of high-level quantum chemistry, but about a million times faster.

Erik Deumens, Senior Director for UFIT Research Computing, highlights the unique ability of HiPerGator to perform large calculations, paving the way for significant scientific breakthroughs. This close collaboration between the university and Ying Zhang's team, responsible for AI support at UFIT, has enabled the acceleration of data analysis, reducing analysis time to just seven hours, down from an initially estimated three days.

This research illustrates the potential of large-scale computer simulations to discover how complex molecules can form from simple building blocks. It marks a step towards understanding the origins of life on Earth and demonstrates the importance of advanced computing infrastructures in contemporary scientific research.
Page generated in 0.277 second(s) - hosted by Contabo
About - Legal Notice - Contact
French version | German version | Spanish version | Portuguese version