Determining how to treat cancer often depends on understanding the genes driving a tumor’s growth. Traditionally, this requires genetic sequencing, a costly process that can take weeks.
Now, researchers at Stanford Medicine have developed an artificial intelligence (AI) tool that predicts gene activity in tumor cells using only standard microscope images of biopsies, potentially saving time and money.
Their findings, published in Nature Communications on November 14, highlight the tool’s potential to transform cancer diagnostics.
The Innovation: SEQUOIA
The new AI program, named SEQUOIA, was created using data from over 7,000 tumor samples across 16 cancer types. It analyzes biopsy images stained with hematoxylin and eosin, a common method used to visualize cancer cells under a microscope.
By studying these images, SEQUOIA predicts which genes in the tumor are active, bypassing the need for traditional genetic sequencing.
According to Olivier Gevaert, Ph.D., senior author of the study, this tool could significantly accelerate clinical decision-making and reduce healthcare costs. “It’s a whole new source of data that we didn’t have before,” Gevaert said.
How SEQUOIA Works
Cancer cells’ gene activity often changes their appearance in subtle ways that are invisible to the human eye but detectable by AI.
SEQUOIA identifies these patterns by integrating data from cancer biopsies, genetic profiles, and images of healthy cells. It can predict the activity of more than 15,000 genes with high accuracy.
In tests, SEQUOIA’s predictions matched actual gene activity with over 80% correlation for some cancer types. The AI performed even better when predicting gene signatures—groups of genes that collectively influence how cancers grow or respond to treatment.
The tool also maps genetic variations across different areas of a tumor, providing a visual representation that helps clinicians understand the tumor’s complexity.
Breast Cancer as a Test Case
To demonstrate its usefulness, researchers focused on breast cancer. They tested SEQUOIA’s ability to replicate results from genomic tests like MammaPrint, which analyzes the activity of 70 genes to assess the risk of cancer recurrence.
SEQUOIA provided similar risk scores based solely on biopsy images, and its predictions aligned with patient outcomes: those identified as high risk had shorter times before recurrence and higher rates of cancer returning.
This success suggests that SEQUOIA could eventually replace expensive gene expression tests for breast cancer and other cancers, offering a faster and more affordable alternative.
Potential Applications and Future Steps
While SEQUOIA has shown promise, it’s not yet ready for clinical use. It will need further testing in clinical trials and approval by the U.S. Food and Drug Administration (FDA). Meanwhile, the Stanford team is improving the algorithm and exploring its use for other cancers.
Gevaert envisions a future where SEQUOIA could analyze gene signatures for any type of cancer, reducing reliance on costly genetic sequencing. This advancement could help personalize cancer treatments and improve outcomes for patients.
Broader Implications
The development of SEQUOIA is part of a larger trend of integrating AI into cancer care. By unlocking information hidden in standard biopsy images, this technology could streamline the diagnostic process, making it more accessible to patients worldwide.
“We’ve shown how useful this could be for breast cancer,” Gevaert said. “The potential to expand this to all cancers is immense.”
If you care about cancer risk, please read studies that exercise may stop cancer in its tracks, and vitamin D can cut cancer death risk.
For more information about cancer, please see recent studies that yogurt and high-fiber diet may cut lung cancer risk, and results showing that new cancer treatment may reawaken the immune system.
The research findings can be found in Nature Communications.
Copyright © 2024 Knowridge Science Report. All rights reserved.
