Artificial intelligence is fundamentally reshaping how scientists decode human DNA and how doctors can personalize health care to match each person's unique genetic makeup. Two major developments announced this week highlight how genomics and AI are moving beyond one-size-fits-all medicine toward truly individualized health strategies, from custom nutrition formulas to targeted disease prevention.

How Is AI Learning to Read the "Hidden" Parts of Our DNA?

For more than two decades since scientists first sequenced the entire human genome, researchers have struggled with a fundamental puzzle: while only 1 to 2 percent of human DNA codes for proteins, the remaining 98 percent contains a mix of evolutionary leftovers and regulatory elements that control when and how genes turn on and off. These non-coding regions hold clues to inherited diseases like cancer, heart disease, and autism, but scientists have lacked the tools to understand which genetic variants actually matter.

Researchers at UC Berkeley have now created a new AI model called GPN-Star that excels at spotting disease-relevant genetic variants hidden within this non-coding DNA. Unlike larger competing models that require massive computing resources, GPN-Star can be trained in just days or even hours using only a handful of processors, making it far more accessible to research teams worldwide.

"Our model excels in making predictions about the pathogenicity of genetic variants, and identifying functional versus non-functional elements in the genome," said Yun Song, a professor of computer science and statistics at UC Berkeley and director of the Berkeley Center for Computational Biology.

Yun Song, Professor of Computer Science and Statistics at UC Berkeley

The key innovation behind GPN-Star is how it learns. Rather than training on individual unaligned genomes from hundreds of thousands of species, which demands enormous computing power, the model uses whole-genome alignments (WGAs) that compare genomes across species to highlight which DNA sequences have been preserved by evolution. This biological shortcut helps the model focus on functional elements instead of wading through junk DNA.

The researchers trained GPN-Star on three different human-anchored alignments representing different evolutionary timescales: one comparing humans to other primates, one to mammals, and one to all vertebrates. They discovered something surprising: models trained at different evolutionary distances excelled at predicting different types of genetic variants. The primate-focused model, for instance, proved better at predicting variants linked to complex traits like schizophrenia risk, which involve thousands of genetic mutations scattered across non-coding regions.

What Does Personalized Health Actually Look Like in Practice?

While GPN-Star represents a breakthrough in understanding genetic disease risk, BGI Group unveiled an entirely different application of genomics and AI: building personalized health management systems that treat each person as biologically unique. At its 27th anniversary event in Shenzhen, the company showcased technologies designed to move health care from reactive treatment to proactive prevention based on individual genetic and metabolic data.

BGI's vision centers on a simple but powerful idea: instead of giving everyone the same health advice, nutrition recommendations, or skincare routines, AI systems can analyze each person's genetic makeup, metabolic profile, and biological markers to create truly customized plans. The company's CEO emphasized that this shift represents "the era of great health and longevity," where the goal is not simply living longer but living better with greater dignity.

"A health checkup is the starting point for improving health, not the end point. Only by continuously accumulating and tracking health data can people understand changes in their bodies earlier and turn health into actions they can take every day," said Yin Ye, CEO of BGI Group.

Yin Ye, CEO of BGI Group

BGI introduced several concrete applications of this personalized approach across different aspects of health and wellness:

• Personalized Nutrition: Instead of taking broad-spectrum supplements without knowing whether you actually need all the ingredients, BGI's system uses whole-genome testing and metabolomic analysis to compare your individual data against longevity baselines from 129 Chinese centenarians and more than 10,000 healthy individuals. Algorithms then generate custom nutrition formulas with precision down to 5 milligrams per ingredient, and can even divide doses based on your biological rhythms.
• Genetic Health Profiling: BGI combines long-read and short-read genome sequencing to build long-term biological profiles for families, covering inherited disease risks, cancer-related risks, complex disease susceptibility, pharmacogenomics (how your genes affect drug response), ancestry, and physical traits.
• Biological Age Assessment: Blood-based epigenetic analysis can examine approximately 28 million DNA methylation sites to assess the biological age of eight major organs and multiple health dimensions, revealing whether your body is aging faster or slower than your chronological age would suggest.

What New Technologies Are Enabling This Shift to Personalized Medicine?

BGI also unveiled several hardware and software innovations that make continuous health monitoring practical for everyday use. An AI-powered multimodal eye imaging system can complete a comprehensive assessment of eye and multi-system health risks in just five to seven minutes by integrating external eye imaging, fundus photography, optical coherence tomography (OCT), and eye-movement tracking.

The company introduced a wearable smart ring that continuously records cardiovascular, sleep, activity, and recovery-related signals through daily use, supporting 24-hour passive monitoring without requiring active user input. This data feeds into personal digital health records that track changes over time.

Perhaps most notably, BGI unveiled the world's first multimodal brain-on-chip system, designed for neuroscience research. For the first time, this platform integrates on-chip cultivation of neuronal cells, neural tissues, and brain organoids alongside electrophysiological and optical functional monitoring on a single device. By enabling the full cycle of cultivation, recording, and modulation on one chip, researchers can conduct long-term, continuous observation and regulation of the same sample, opening new possibilities for studying brain organoid development, neurological disease modeling, and neuropharmacological screening.

How Can Researchers and Clinicians Use These AI Tools?

The practical implications of these advances extend beyond individual consumers to the broader scientific and medical communities. The UC Berkeley team has published genome-wide predictions from GPN-Star that highlight genetic variants most likely to influence inherited traits, giving biologists a roadmap for prioritizing experiments.

• Experimental Prioritization: Because researchers cannot experimentally test every single genetic variant in the human genome, GPN-Star's predictions help scientists focus their lab work on variants most likely to have the greatest impact on human health and disease.
• Model Accessibility: Unlike massive competing models that require thousands of powerful processors and months of training time, GPN-Star can be modified, adapted, and improved by research teams worldwide because it requires minimal computational resources, accelerating global progress in understanding genetics.
• Clinical Integration: BGI's personalized health systems are designed to integrate multi-source physical examination data, wearable signals, and genetic information into unified health frameworks that support continuous management rather than one-time testing, enabling doctors to track whether lifestyle changes are actually working.

The convergence of these two developments, one focused on understanding genetic disease mechanisms and the other on practical personalized health management, reflects a broader transformation in life sciences. Rather than treating everyone according to population averages, AI and genomics are making it possible to identify individual differences first, then support more precise decisions in research, health management, nutrition, and disease prevention.

BGI's chairman emphasized this vision, stating that the company will remain guided by health, supported by more advanced and more accessible life science tools and genomic foundation models, continuing to promote the deep integration of life sciences and clinical medicine. As these technologies mature and become more widely available, the era of truly personalized medicine, informed by AI and genomics, appears to be arriving faster than many expected.