Glowing “Fingerprint” from a Blood Test May Find Ovarian Cancer Early

Machine learning and tiny glowing nanosensors enable the detection of ovarian cancer from a blood test.

Grantee: Daniel Heller, PhD
Institution: Memorial Sloan Kettering Cancer Center in New York, NY
Area of Focus: Cancer Drug Discovery
Grant Term: 7/1/2018-6/30/2022

“Ovarian cancer is very hard to detect, but if it can be caught early, it is much easier to treat. My lab found that carbon nanotubes — tiny fluorescent particles — can respond to proteins in the blood and give a diverse set of signals. We used machine learning algorithms to make sense of these signals. We found that our nanosensors  can be used to detect ovarian cancer better than conventional markers, and they may be able to be adapted to detect many other cancers.”  –Daniel Heller, PhD

The Challenge:  When ovarian cancer is found before it has spread, more than 90% of women live longer than 5 years after their diagnosis. The problem is, so far, there is no effective way to screen for ovarian cancer in its early states. 

Doctors may suggest that women with a higher risk for developing ovarian cancer due to inherited genetic factors have a pelvic exam, transvaginal ultrasound, and blood test to check for the CA125 tumor marker (also known as a biomarker). This strategy, though, has not been proven to reduce deaths caused by ovarian cancer, and it results in many false positives.

Currently, almost 60% of cases of ovarian cancer are diagnosed when they’ve already spread to organs that aren’t near the ovaries. At that stage, the 5-year survival rate is 30%.

Discovering a way to find ovarian cancer earlier would dramatically improve the outcomes for people with this type of cancer.

The Research: Biomedical engineer and American Cancer Society research grantee Daniel Heller, PhD, leads the Cancer Nanomedicine Laboratory at Memorial Sloan Kettering Cancer Center. He and his lab team are working to develop a way to find ovarian cancer in its early stages and recently published their discovery of ovarian cancer “spectral fingerprints.”

INSIDE DANIEL HELLER’S LAB

Small nanotubes light up the screen in Dr. Heller’s lab. Video credit: Molecular Pharmacology Program, Memorial Sloan Kettering

Heller’s lab developed multiple nanoscale sensors, also called nanosensors, to collect these “fingerprints” from blood samples from people with different types of ovarian cancer and non-cancerous conditions. Nanoscale means the technology is small— really small. The carbon nanotubes that make up the sensors are nearly 100,000 times smaller than the thickness of a strand of human hair.

The researchers designed each sensor so that it can detect many different molecules in a blood sample. With an array of multiple nanosensors, the combined responses emit a distinct color and pattern of light when they sense certain biomarkers in their environment—like a fingerprint.

To identify which glowing fingerprint corresponds with a type of cancer or with a noncancerous condition, like endometriosis, the team uses machine learning, a type of artificial intelligence. Heller’s researchers found that their nanosensor detected ovarian cancer more accurately than the currently available blood biomarker tests.

Why Does It Matter? Heller’s combined approach of using nanosensors and machine learning detects high-grade serous ovarian carcinoma, the deadliest form of ovarian cancer, with greater sensitivity and specificity than current blood-based biomarkers.

The hope for patients is that researchers can develop the technology further so that doctors can eventually use it to rapidly screen for early-stage ovarian cancer and many other cancers. As noted in the recent publication, Heller’s “sensor technology has the potential to be developed into an inexpensive and rapid screening tool that produces a single, easy-to-interpret test result in primary-care settings.”

What Is Machine Learning

Machine learning uses computer systems that are able to learn and adapt without having a set of explicit instructions to follow. The way they learn is similar to how people solve problems, by analyzing and drawing inferences from patterns in data.

In some cases machine learning can gain insight or automate decision making when people would not be able to —especially not at the same speed and scale.

Familiar results of machine learning include chatbots, predictive text, and how browser search responses and social media feeds are presented.

In medicine, machine learning programs can be trained to look at medical information and pictures to look for established markers of disease. For instance, machine learning could be used to develop a tool to predict the risk of developing cancer based on a mammogram.