Together, machine learning and tumor DNA provide new tools for colorectal cancer patients

By Samantha Black, PhD, ScienceBoard editor in chief

January 2, 2020 -- Researchers utilize a new machine learning platform to identify patients with colorectal cancers and predict their disease severity and survival. The study, published in Science Translational Medicine on January 1, involves samples from over a thousand patients.

Colorectal cancer is the third most deadly cancer in the world. According to the American Cancer Society, there were over 145,000 cases in the United States in 2019 alone. Early detection is essential to improving survival in patients with these cancers. There has been limited success in developing effective, noninvasive diagnostic approaches for colorectal cancers. Circulating tumor DNA (ctDNA) is a powerful diagnostic and prognostic biomarker in many cancers.

Therefore, the researchers wanted to explore the predictive diagnostic power of a machine learning algorithm using cell-free DNA (cfDNA). cfDNA methylation data were randomly divided into training and validation datasets to build diagnostic and prognostic models. Machine learning was applied to markers to that researchers indemnified in the study and models were evaluated with cross-validation and ROC methods.

Since DNA methylation is a key epigenetic process driving tumor growth, methylation profiles using cfDNA can be a powerful diagnostic tool in some cancers. The researchers developed cfDNA methylation markers for colorectal cancer. Next, they determined the efficacy of CpG markers (common methylation sites in cancer mutations) in screening for colorectal cancers. Prediction models were constructed with nine selected methylation markers.

The diagnostic model in this study accurately discriminated patients with colorectal cancer from normal individuals using a cd-score. The researchers suggest that the methylation markers are associated with carcinogenesis and development of colorectal cancers. This model accurately distinguished patients from healthy individuals (via cg10673833 methylation) with a sensitivity and specificity of 87.5% and 89.9%, respectively, and outperformed a clinically available blood test named CEA, the only global blood test for the disease.

The prognostic model using a five-marker panel was also constructed. The cp-score predicted by the model could distinguish patients with colorectal cancer with different prognoses which helped practitioners identify patients who would need more aggressive treatment and monitoring. It should be noted that this study was limited by a short clinical follow-up time of only around 26 months and further studies with longer surveillance periods are needed.

The researchers found two molecular subgroups based of cfDNA methylation. Cluster 1 tumors were frequently observed in females with left-sided lesions. While cluster 2 showed significantly poorer survival rates indicative of stage III and IV.

Overall, the researchers showed the usefulness of using cfDNA methylation markers for diagnosis, prognostication and surveillance of colorectal cancer. Next steps include validation of clinical applicability with large-scale randomized clinical trials.

Do you have a unique perspective on your research related to genetics and/or cancer biology? Contact the editor today to learn more.


Join The Science Advisory Board today!

New microscopy technique helps scientists understand cancer growth and spread
Researchers from Duke Cancer Institute have developed a new technique to observe stem cell mutations occurring in colon cancer.
Scientists elucidate the genetic impact of host immune system on the microbiome
Researchers at the University of Chicago conducted research that suggests that the genetic differences in the immune system contribute to bacterial colonization...
Global Collaboration Launched for Large-Scale DNA Study Exploring Connection Between Genetics and Disease
A new partnership was announced between Intermountain Healthcare and deCODE Genetics to collaborate on the largest DNA mapping effort undertaken in the...

Copyright © 2020

Science Advisory Board on LinkedIn
Science Advisory Board on Facebook
Science Advisory Board on Twitter