image: Figure 2. Overview of SAMP-Score model development. Each model was assessed according to a range of criteria, which are visualised in the model metrics (Figure 3A) and confusion matrix heatmaps (Figure 3B), as well as neural network map (Figure 3D) and ROC curves (Figure 3E; Supplementary Figure 4). The model metrics are nuanced and can be misleading when viewed in isolation. For instance, accuracy is a measure of correct predictions and is often relied upon as a single readout of model performance. But in a hypothetical example where there are 99 majority cases (e.g., NonSen) and 1 minority case (e.g., Sen) then a model may be 99% accurate by simply always predicting NonSen; but this would not be a useful tool. Therefore, particularly in senescence research where instances are likely to be imbalanced, particular care in assessing model performance must be taken.
Credit: Copyright: © 2025 Wallis et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
“Overall, this study further demonstrates the utility of high-content morphological analysis as a tool for the identification of senescent cells.”
BUFFALO, NY — December 1, 2025 — A new research paper featured on the cover of Volume 17, Issue 11 of Aging-US was published on October 30, 2025, titled “SAMP-Score: a morphology-based machine learning classification method for screening pro-senescence compounds in p16 positive cancer cells.”
In this study led by first author Ryan Wallis along with corresponding author Cleo L. Bishop, from Queen Mary University of London, researchers developed a machine learning tool to identify compounds that induce cancer cells into senescence. The tool, called SAMP-Score, offers a new strategy for drug discovery in cancers with poor treatment options like basal-like breast cancer.
Senescence is a process where damaged or aged cells stop dividing. In cancer therapy, inducing senescence is an approach to control tumor growth. However, it is difficult to detect true senescence in cancer cells that already appear aged. These cancers, often called Sen-Mark+ cancers, include basal-like breast cancer and typically lack reliable markers to confirm senescence. SAMP-Score was designed to address this problem.
Instead of relying on traditional markers, the researchers built a machine learning model trained to recognize patterns based on senescent cells’ shape and structure under a microscope. These visual patterns, known as senescence-associated morphological profiles (SAMPs), allowed the model to distinguish real signs of aging from other effects such as toxicity or normal variation. By analyzing thousands of cell images, the model learned to classify whether a cell had truly entered senescence.
“To demonstrate the potential application of SAMP-Score in p16 positive cancer therapeutic discovery, we assessed a diversity screen of 10,000 novel chemical entities in MB-468 cells (p16 positive BLBC).”
The team used SAMP-Score to screen more than 10,000 experimental compounds. One compound, QM5928, consistently triggered senescence in several cancer cell types without killing them, making it a promising candidate for further study. Importantly, it worked in cancers resistant to known drugs like palbociclib, which are often ineffective in cancers with high p16 expression like basal-like breast cancer.
Further analysis revealed that QM5928 caused the p16 protein to move into the nucleus of cancer cells, a possible sign that the protein is helping stop cell division. This subtle effect was only detectable using the detailed imaging and analysis made possible by SAMP-Score, highlighting the tool’s ability to distinguish true senescence from toxic responses and making it a powerful resource in cancer drug discovery.
By combining machine learning with high-resolution imaging, this study introduces a new way to find and evaluate cancer therapies. SAMP-Score could accelerate efforts to develop treatments that exploit the body’s natural aging processes to fight cancer, especially for patients with resistant tumors. The tool is openly available at GitHub, making it accessible for other researchers exploring senescence-based cancer therapies.
DOI: https://doi.org/10.18632/aging.206333
Corresponding author: Cleo L. Bishop – c.l.bishop@qmul.ac.uk
Abstract video: https://www.youtube.com/watch?v=qXI_KI3EgHE
Keywords: aging, SAMP-Score, senescence, senescent marker positive cancer cells, Sen-Mark+, machine learning, pro-senescence, high-throughput compound screening
Click here to sign up for free Altmetric alerts about this article.
______
To learn more about the journal, please visit www.Aging-US.com and connect with us on social media:
- X
- YouTube
- Bluesky
- Spotify, and available wherever you listen to podcasts
Click here to subscribe to Aging-US publication updates.
For media inquiries, please contact media@impactjournals.com.
Journal
Aging-US
Method of Research
News article
Subject of Research
Not applicable
Article Title
SAMP-Score: a morphology-based machine learning classification method for screening pro-senescence compounds in p16 positive cancer cells
Article Publication Date
30-Oct-2025
COI Statement
CLB has acted as a consultant for Senisca, StarkAge and AKLRD. CLB’s laboratory received funding from ValiRx.