AI learns how to predict breast cancer risk from single cell pressure responses

A study led by City of Hope and UC Berkeley researchers found that the physical and mechanical properties of normal human breast epithelial cells can provide a “functional readout” of biological age and breast cancer susceptibility.

The research team has created a new high-throughput microfluidic platform that can assess women’s breast cancer risk at the cellular level. The mechano-node pore-sensing (mechano-NPS) platform, which researchers claim is the first of its kind, compresses individual breast epithelial cells, creating a taxing environment to measure how the cells deform, recover, and behave under stress.

Using this platform, researchers discovered the unexpected insight that breast cells appear to have a “mechanical age” independent of a person’s chronological age, demonstrated by how the cells physically respond to stress. For the study, the team developed MechanoAge, a machine learning classifier for estimating chronological age based on mechanical phenotypes, and Mechano-RISQ, a biological age-based risk index.

“We found that the higher the mechanical age, which is determined by how a cell responds to compression through a microfluidic device, the higher the risk of breast cancer,” explained Lydia Thorne, Ph.D., the Almy C. Maynard and Agnes Offield Maynard Professor of Mechanical Engineering at the University of California, Berkeley. The researchers suggest their innovative approach fills a critical gap in risk assessment and has the potential to save countless lives, as more than 90% of women have no genetic predisposition or family history of breast cancer.

Sohn is co-senior author on a paper the team published in 2016. e-biomedicine“A machine learning platform that identifies individuals susceptible to breast cancer based on the mechanical properties of single cells. MechanoAge concludes, “Aging-associated biomechanical changes may be fundamental features of cellular function, with distinct mechanistic phenotypes underlying important processes in aging, cancer, and potentially other diseases. Recognizing and exploiting these biomechanical markers has the potential to greatly enhance early detection, improve risk stratification, and improve targeted intervention strategies.”

Breast cancer is one of the most frequently diagnosed cancers worldwide and the leading cause of cancer-related deaths in women, and “efforts to improve risk stratification and early detection strategies have long been underway,” the authors noted.

Approximately 6% of women who develop breast cancer have a known genetic mutation. However, for women outside this group, risk is estimated indirectly based on population models and measurements such as breast density. These approaches can overestimate or underestimate a woman’s individual breast cancer risk, leading to over- or under-screening, unnecessary worry, and missed warning signs. And despite significant advances in screening technologies and therapeutic interventions, determining precisely which individuals are most likely to develop breast cancer, especially those considered to be at average risk, remains what the research team calls “one of the most enduring challenges in oncology and public health.”

The researchers added that for these “ostensibly average-risk individuals,” it “remains difficult to identify those with potential risks due to cellular, molecular, and biophysical changes that current models are not designed to capture.”

Currently, there are no non-genetic tests that can identify women who are at increased risk for breast cancer. The disadvantage of mammograms is that they can only detect cancer after it has begun to grow. “For women with known genetic risk factors for breast cancer, there are things you can do, such as following higher-risk screening protocols,” said co-senior author Dr. Mark Laberge, professor in City of Hope’s Department of Population Sciences.

Emerging evidence is emerging linking cellular aging and biophysical changes to cancer susceptibility. In the reported study, researchers used the mechano-NPS platform to profile primary human breast epithelial cells (HMECs) from women of various ages and risk backgrounds. We also developed a machine learning algorithm that identifies and measures cells that show signs of accelerated aging and quantifies individual breast cancer risk scores.

This type of mechanonode pore sensing measures current flowing through a liquid-filled channel in the same way that current flowing through a wire is measured. Once the cell passes, the current is interrupted and measurements are generated about the cell’s size and shape. The researchers squeeze the cells by making part of the channel very narrow and measure the time it takes for each cell to regain its normal shape.

The researchers used a machine learning algorithm developed to detect differences between cells in older and younger women. Researchers have discovered that the physical properties of breast cells change with age. Older women’s cells were harder and took longer to recover after being compressed.

Then, a surprising discovery was made. Some young women had cells that behaved as if they had come from older women. These cells were taken from women who carry genetic mutations that put them at higher risk for breast cancer. The researchers then refined an algorithm that assigned a risk score based on all the mechanical and physical properties measured within the cells. The algorithm was successful in identifying women with known genetic risks. The researchers then used this to compare cells from the healthy breasts of healthy women, women with a family history of breast cancer, and women with breast cancer in the other breast. The researchers state that “normal epithelial cells from women with germline mutations, a strong family history of cancer, or contralateral breast cancer exhibit a mechanically aged phenotype despite normal histology.” “Together with previous molecular and epigenetic studies, these findings support the model that accelerated biological aging of the breast epithelium may underpin breast cancer susceptibility across genetic and non-genetic risk groups.”

The researchers used the MechanoAge platform to move the science to the cellular level, calculating risk by looking for physical changes in individual cells. “Mechanical phenotyping captures an integrated cellular state that reflects the underlying molecular network rather than a single biomarker,” the researchers said. “Mechano-RISQ provides a proof-of-principle approach to identify individuals at high risk for breast cancer, particularly within average-risk populations, and has the potential to complement existing risk models by incorporating biophysical measurements of breast epithelial cell aging.”

“We were able to see exactly which women were at high risk for breast cancer and which were not,” Laberge said. “By converting physical changes in cells into quantifiable data, this tool gives women something concrete to discuss with their doctors. It provides not only an estimate of risk, but also evidence extracted directly from their own cells.” In the paper, the scientists go on to say, “This approach may allow for earlier and personalized risk stratification, particularly for women who have no identifiable high-risk mutations but have tissue conditions that make them susceptible.”

Importantly, the AI ​​platform uses simple electronic equipment that is easy and affordable to replicate at scale. “Our team is not the first to measure the mechanical properties of cells, but other approaches require advanced imaging techniques that are expensive, cumbersome, and of limited availability,” Thorne said. “In contrast, MechanoAge uses a simpler computer chip than the Apple Watch and ‘Radio Shack parts’ that are cheaper and easier to assemble, making the device potentially much more scalable.”

Engineers study the aging of materials such as metals, concrete, and polymers, but this is the first time mechanical aging has been quantified in biological cells. The discovery that cells have a “mechanical age” independent of an individual’s chronological age would not have been possible without MechanoAge.

This research stems from more than 12 years of collaboration between the two laboratories and combines engineering innovation with cancer and aging biology. The long-term partnership has enabled discoveries that neither group could have reached alone. “This is a true collaboration. We learned a lot from each other,” Song said. “In my opinion, this is what happens in true collaborations that develop over a long period of time,” Laberge added. “This result is different from what we originally imagined.”