Researchers have developed an artificial intelligence (AI)-powered computational program that may be capable of predicting the activity of thousands of genes within tumor cells based on standard microscopy images of a biopsy, according to a recent study published by Pizurica et al in Nature Communications. The findings demonstrated that the novel tool may use routinely collected biopsy images to predict genetic variations in breast cancers as well as patient outcomes.
Background
To determine the type and severity of a cancer, pathologists typically analyze thin slices of a tumor biopsy under a microscope. But to figure out what genomic changes may be driving the tumor’s growth—information that can guide treatment decision-making—physicians must perform genetic sequencing of the RNA isolated from the tumor, a process that can take weeks and costs thousands of dollars.
Physicians have increasingly guided the selection of which cancer treatments to recommend to a patient based on not only which organ a patient’s cancer affects, but which genes a tumor is using to fuel its growth and spread. ‘Turning on’ or ‘off’ certain genes could make a tumor more aggressive, more likely to metastasize, or more or less likely to respond to certain drugs. However, accessing this information often requires costly and time-consuming genomic sequencing.
Physicians don’t often look at genes one at a time to make clinical decisions. Instead, they examine gene signatures that include hundreds of different genes. For instance, many cancer cells activate the same groups of hundreds of genes related to inflammation or hundreds of genes related to cell growth.
Prior research has shown that the gene activity within individual cells may alter the appearance of those cells in ways that are often imperceptible to a human eye. As a result, the researchers turned to AI to find these patterns.
Study Methods and Results
In the recent study, the researchers developed the SEQUOIA AI-powered tool using data from 7,584 diverse tumor samples from 16 different cancer types. Each biopsy had been sliced into thin sections and prepared using hematoxylin and eosin staining. Information on the cancers’ transcriptomes was also available.
After the researchers integrated their new cancer biopsies and other data sets—including transcriptomic data and images from thousands of healthy cells—the AI program was able to predict the expression patterns of more than 15,000 different genes from the stained images. For some cancer types, the AI-predicted gene activity had more than an 80% correlation with the real gene activity data. In general, the more samples of any given cancer type that were included in the initial data, the better the model performed on that cancer type.
“It took a number of iterations of the model for it to get to the point where we were happy with the performance, [b]ut ultimately for some tumor types, it got to a level that it can be useful in the clinic,” explained senior study author Olivier Gevaert, PhD, Professor of Biomedical Data Science at Stanford University.
Compared with its performance at predicting individual gene expression, the SEQUOIA tool was even more accurate at predicting whether such large genomic programs were activated. To make the data accessible and easy to interpret, the researchers programmed the tool to display the genetic findings as a visual map of the tumor biopsy, letting researchers and clinicians see how genetic variations might be distinct in different areas of a tumor.
To test the utility of SEQUOIA for clinical decision-making, the researchers identified breast cancer genes that the model could accurately predict the expression of and that are already used in commercial breast cancer genomic tests. For example, the U.S. Food and Drug Administration (FDA)-approved MammaPrint test analyzes the levels of 70 breast cancer–related genes to provide patients with a score of the risk their cancer is likely to recur.
“Breast cancer has a number of very well-studied gene signatures that have been shown over the past decade to be highly correlated with treatment responses and patient outcomes,” Dr. Gevaert indicated. “This made it an ideal test case for our model,” he added.
The researchers discovered that the SEQUOIA tool could provide the same type of genomic risk score as the MammaPrint tool using only stained images of tumor biopsies. The results were repeated on multiple different groups of patients with breast cancer. In each case, patients identified as high risk by the SEQUOIA tool experienced poorer outcomes, with higher rates of cancer recurrence and a shorter time before their cancer recurred.
Conclusions
“This kind of software could be used to quickly identify gene signatures in patients’ tumors, speeding up clinical decision-making and saving the health-care system thousands of dollars,” highlighted Dr. Gevaert.
The novel AI model cannot yet be used in a clinical setting, since it requires further assessment in clinical trials and FDA approval before it’s used in guiding treatment decisions. Nonetheless, the researchers are currently working to improve the algorithm and study its potential applications. In the future, the SEQUOIA tool may reduce the need for expensive gene-expression tests.
“We’ve shown how useful this could be for breast cancer, and we can now use it for all cancers and look at any gene signature that is out there,” Dr. Gevaert underscored. “It’s a whole new source of data that we didn’t have before,” he concluded.
Disclosure: The research in this study was funded by the National Cancer Institute, a fellowship of the Belgian American Educational Foundation, Fonds Wetenschappelijk Onderzoek-Vlaanderen, the Fulbright Spanish Commission, and Ghent University. For full disclosures of the study authors, visit nature.com.
