Researchers at Duke University School of Medicine have developed a powerful new tool that can detect cells that play a significant role in aging, known as senescent cells, with remarkable accuracy across a wide range of tissues and technologies.

This AI-based method, called DeepScence, could help scientists better understand aging and disease, and potentially guide future treatments.

Senescent cells are cells that have stopped dividing but don’t die off as they should. Instead, they linger in the body and can contribute to inflammation, tissue damage, and diseases like cancer and fibrosis.

“DeepScence represents a major step forward in our ability to study senescence at the single-cell level,” said Zhicheng Ji, PhD, assistant professor of biostatistics and bioinformatics and lead author of the study.

The researchers outlined their discoveries about DeepScence in Cell Genomics in October 2025. 

The tool uses deep learning to spot senescent cells in both single-cell and spatial transcriptomics data types of genetic data that show which genes are active in individual cells and where those cells are located in tissue. It works across different species, cell types, and experimental platforms, outperforming existing methods.

It is powered by two key innovations. The first, CoreScence, is a carefully curated set of genes that represent a consensus of known senescence markers. The second is an autoencoder model, which is a type of AI that transforms gene signals into a continuous score and a clear yes-or-no answer about whether a cell is senescent.

Before DeepScence, researchers lacked a reliable way to detect senescent cells across diverse datasets. This made it hard to study where these cells appear, how many there are, and what role they play in health and disease.

DeepScence fills that gap, offering a unified, validated approach that can be used in lab experiments, animal models, and even human tissue. It allows researchers to map senescent cells in both diseased and healthy tissues more precisely, improving understanding of aging and disease progression and helping to identify new drug targets. The tool could guide the use of senolytics drugs that remove senescent cells by showing where and when they should be used.

Because DeepScence includes spatial data, it can also show where senescent cells cluster in tissues, such as near immune cells or structural support cells, Ji said. This could help predict disease outcomes and tailor treatments to individual patients.

Next, the team plans to test DeepScence in more species, tissues, and diseases, including clinical samples, to confirm its accuracy and broaden its use. They also aim to link DeepScence’s scores with treatment outcomes, helping determine whether senescent cell patterns can predict how patients respond to therapies.

Source: https://biostat.duke.edu/news/new-duke-discovery-could-help-target-cells-drive-aging-and-disease