In certain cancer cases, pinpointing the exact origin of the cancer is challenging, which complicates treatment decisions. Specific cancer treatments are tailored to different cancer types, and not knowing where the cancer originated from hinders treatment options.

Fortunately, a collaboration between researchers at MIT and the Dana-Farber Cancer Institute has unveiled a machine-learning model, named OncoNPC, that can predict a tumor’s origin by analyzing the genetic sequence of about 400 genes.

Upon testing, the model showcased an impressive ability to correctly identify the origins of tumors 80% of the time. For tumors where predictions were made with high confidence, the accuracy surged to about 95%. This is monumental for cases classified as cancers of unknown primary (CUP), representing 3 to 5 percent of all cancer patients.

Utilizing the OncoNPC model on about 900 CUP tumors revealed that it could make high-confidence predictions for 40% of them. Further validating the model’s precision, the predicted cancer types corresponded with patients’ survival times, aligning with known prognoses for those cancer types.

Additionally, the model’s implementation could have reshaped treatment decisions for patients. A subset of CUP patients who received treatment congruent with the model’s predictions had more favorable outcomes compared to those who received mismatched treatments.

The researchers emphasized that a significant portion of patients could benefit from existing precision treatments if the cancer’s origin was identified through their model.

The team, funded by multiple institutions, including the National Institutes of Health, is optimistic about enhancing the model’s capabilities. Incorporating other data types like radiology and pathology images aims to equip the model to offer comprehensive insights, potentially guiding optimal treatment options.

To view the original source, visit: MIT News  


RadParts, a TTG Imaging Solutions Company, is the world’s largest independent distributor of OEM replacement parts. We specialize in low-cost parts for repairing linear accelerators and radiation equipment. Our mission is to provide high-quality, user-friendly, low-cost components and support for linear accelerators and radiation equipment. Contact RadParts at 877-704-3838 to learn more.

Written by the digital marketing team at Creative Programs & Systems: