A Yale research team has created a new imaging technique that reveals the hidden connections between aging, disease, and genetic activity in human cells. 

Using a novel machine learning approach, the team found that tissue samples, under a microscope, can reveal genetic variants, gene activity, and even estimates of a person’s age. 

“Our study shows that ordinary tissue images contain patterns that can reliably predict gene expression and reveal a person’s age information that was previously hidden to the naked eye,” said study lead author Ran Meng, a postdoctoral researcher in Yale’s Department of Molecular Biophysics and Biochemistry and program in computational biology and biomedical informatics.

“The improved image quality let us link genetic features.” Meng added. “And the models are able to crunch a great amount of data very accurately and shine a light on the image regions that push the prediction toward an older or younger age.”

The new technique could lead to the development of better diagnostic practices using routine pathology slides, and the ability to predict disease risk by spotting abnormal tissue patterns early. The research is published in the Proceedings of the National Academy of Sciences. 

“One key aspect of genetics is the genotype-phenotype connection,” said co-author Mark Gerstein, the Albert L. Williams Professor of Biomedical Informatics at Yale School of Medicine who is also a professor of molecular biophysics and biochemistry; of computer science; and of statistics and data science in Yale’s Faculty of Arts and Sciences.

A genotype refers to an organism’s genetic makeup, while a phenotype refers to observable characteristics which are shaped by genotype and environmental factors. Examples would include traits like height and eye color, or external conditions like adequate nutrition and clean water, as well as more complex behavioral traits or the presence of disease.

“One of the new research frontiers is ‘multi-modality’ connecting genotype to all sorts of other types of data that describe phenotype,” Gerstein said. “In this paper, we make an advance in connecting genotype to image features.”

For the study, researchers applied machine learning techniques to analyze tissue images taken from healthy human donors, which enabled them to unlock hidden signs of aging and gene activity that appear in human cells. Cell appearance is tied to both genes and the aging processes.

Using histology slides, genetic information, and RNA data from 838 donors covering 12 different tissue types and more than 10,000 images the researchers built computer models that could spot genetic variants linked to tissue appearance. The models were also able to predict gene expression when genes are switched “on” or “off” and a person’s age.

One of the machine learning models, they said, can predict gene expression based on an image, with some tissue samples lung, heart, and testis showing particularly strong predictive accuracy.

Another model could estimate a person’s chronological age from tissue samples, with skin, tibial nerve, tibial artery, and testis tissue giving the most accurate age predictions, because they show more age-related changes.

Overall, researchers found that the shape, size, and structure of cell nuclei carry a lot of biological information. They found 906 points in the human genome that were strongly tied to nuclei appearance in different tissues. There were also strong connections between nuclear shape and gene activity.

Source: https://news.yale.edu/2025/11/18/uncovering-hidden-cellular-connections-bridge-aging-and-disease