Scientists at the Georgia Tech Integrated Cancer Research Center have developed a new early diagnostic test for ovarian cancer using a combination of machine learning and information on blood metabolites. The test demonstrated a 93% accuracy rate in detecting ovarian cancer among samples from the study group. The findings and methodology were published in the journal Gynecologic Oncology.
John McDonald, professor emeritus in the School of Biological Sciences and founding director of the ICRC, explained that the new test outperforms existing tests in detecting ovarian cancer, particularly in the early stages of the disease. The researchers used computer models to create a personalized approach to ovarian cancer diagnosis, using a patient’s individual metabolic profile to determine the probability of the presence or absence of the disease. McDonald stated that this personalized, probabilistic approach provides more clinically informative and accurate results compared to traditional binary tests.
Ovarian cancer is often referred to as the “silent killer” because it is typically asymptomatic in the early stages and difficult to detect until it has advanced. The average five-year survival rate for late-stage ovarian cancer patients, even after treatment, is only around 31%. However, if ovarian cancer is detected and treated early, the average five-year survival rate is more than 90%.
McDonald emphasized the urgent need for an accurate early diagnostic test for ovarian cancer, stating that the development of such a test has been pursued for over three decades with little success. Due to the molecular heterogeneity among patients, it has been challenging to identify a universal diagnostic biomarker for ovarian cancer. To overcome this obstacle, the researchers turned to machine learning as an alternative approach.
Dongjo Ban, a co-author of the study, explained that they focused on metabolic profiles as the backbone of their analysis, as changes in metabolites reflect underlying changes at the molecular level. The team used mass spectrometry to identify the presence of metabolites in the blood, and machine learning algorithms to build predictive models based on the detected metabolites. By incorporating the presence of different metabolites in the blood as features in their models, the researchers were able to establish an accurate diagnostic test for ovarian cancer.
It should be noted that less than 7% of the metabolites circulating in human blood have been chemically characterized, making it impossible to pinpoint the specific molecular processes contributing to an individual’s metabolic profile. However, the study demonstrates that even without this knowledge, the presence of different metabolites can still be utilized in the development of accurate diagnostic models.
Jeffrey Skolnick, another co-author of the study, explained that thousands of metabolites are known to be circulating in the human bloodstream, and with the use of mass spectrometry and machine learning, an accurate ovarian cancer diagnostic can be established. The researchers believe that this breakthrough in early detection of ovarian cancer could also pave the way for similar advancements in the detection of other types of cancers.