The mystery of how we became human still drives scientific inquiry, especially among researchers probing the ancient genetic shifts that gave rise to our complex brains, capacity for language, and upright posture—traits that set us apart from our closest ape relatives.
“What we found in our study was that a range of different traits—skeletal, neuropsychiatric, pigmentation, cholesterol synthesis, and so on—were accelerated at different points in the history of humans,” said Vagheesh Narasimhan, who co-authored a study published in Cell Genomics in January 2025. Narasimhan is an assistant professor in the College of Natural Sciences at The University of Texas at Austin (UT Austin).
Vagheesh Narasimhan, UT Austin. (Credit: UT Austin)
The study integrated three powerful data sources: ancient DNA from fossils; 3D MRI scans from hundreds of thousands of participants in the UK Biobank revealing the structure of the brain, skeleton, and major organs; and comparative functional genomics that mapped how the human genome aligns—and diverges—from that of chimpanzees, orangutans, and other great apes. By layering these datasets, researchers were able to uncover where bursts of human-specific evolutionary changes and genetic mutations likely occurred.
“We looked at gene expression and gene regulation through embryo development between humans and other primates, particularly Rhesus macaque,” Narasimhan said. “We then carried out genomic enrichment analysis, which determines whether the overlap between our evolutionary annotation and our annotation associated with traits is more than we expect by chance compared to the genome wide average.”
Frontera (top), Lonestar6 (bottom left), Corral (bottom right) are strategic national computing supercomputing resources at the Texas Advanced Computing Center. (Credit: Jorge Salazar, TACC)
Narasimhan and colleagues leveraged this method to look at whether sections of the human genome associated with traits had bursts at particular time intervals.
With advanced computing power from the Texas Advanced Computing Center (TACC), scientists were able to identify when key human traits may have undergone major evolutionary changes. TACC supported the research by awarding Narasimhan allocations on its Frontera and Lonestar6 supercomputers, along with data storage and management resources on the Corral system.
Lonestar6 helped the researchers process 80,000 3D MRI images of the heart, brain, liver, and pancreas, as well as hip, knee, spine, and whole-body X-ray scans from the UK Biobank.
“We trained AI models for segmentation and classification on the imaging data using TACC GPU (graphics processing unit) resources, particularly Lonestar6, which has a large number of GPUs that were capable of processing this type of data,” Narasimhan said.
Major time points of primate evolution relevant in this study are highlighted. Genomic annotations corresponding to evolutionary time periods are shown in color on the timeline. (Credit: DOI:10.1016/j.xgen.2024.100740)
“For carrying out genomic analyses, we are heavy users of the CPU (central processing unit) infrastructure on Frontera, largely because the genome is a very large data problem,” he added. “Having a large number of CPU nodes on a supercomputing cluster like Frontera was tremendously useful to shrink compute time from a linear process to a parallel process and allow the study to happen.”
The HIPAA protections on TACC’s Corral data storage allowed Narasimhan to simultaneously compute on two different environments, Lonestar6’s GPUs and Frontera’s CPUs.
“It’s impossible to do this work without this integrated enterprise at TACC,” Narasimhan said.
Narasimhan is excited about the new GPU computing capacity with TACC’s AI-focused Vista supercomputer, which entered production in November 2024.
“We’re hoping to use Vista soon and continue our work,” he said. “TACC’s vision to keep pace with new data generation is transformative.”
A more recent study by Narasimhan published in April 2025 in the journal Science also acknowledges TACC support. It found genetic correlations between pelvic proportions and traits such as osteoarthritis, walking speed, and back pain, giving insight into facets of the obstetrical dilemma—the biological tradeoffs between the size of a mother’s birth canal and the brain of her child.
“To truly understand change in the human genome, we need massive amounts of data from a vast number of individuals to look at what’s happening in each of these three billion DNA bases,” Narasimhan said. “It’s a monumental data problem that can only be tackled by using supercomputing infrastructure.”
This article was originally published by TACC and is reprinted with permission.
