AI Tool OncoNPC Aids in Identifying Cancer Origin for Improved Treatment Options
In a study published in Nature Medicine, researchers at Dana-Farber Cancer Institute have developed an AI model named OncoNPC, short for Oncology NGS-based Primary cancer type Classifier, that utilizes gene sequencing data to predict the primary source of a patient’s cancer. This tool could significantly aid physicians in treating cancers that are traditionally difficult to diagnose.
- OncoNPC was trained on medical records from 36,445 patients with known primary tumors.
- The model accurately predicted the origin of about 80% of tumors with known types, and made high-confidence predictions in 65% of the tumors, which were 95% accurate.
- When applied to 971 cases of cancers of unknown primary (CUP), OncoNPC predicted the tumor’s origin with high confidence for 400 cases (41.2%).
- The model is interpretable, providing transparency in its predictions, which could help clinicians trust the tool.
- Patients with CUP who received treatments matching the OncoNPC predictions had longer survival rates and were 2.2 times more likely to be matched to approved targeted medicines.
- The tool has only been studied using retrospective data and needs to be tested in a clinical trial for further validation.
- OncoNPC shows promise in aiding the diagnosis and treatment of cancers, particularly for cases where the primary source is unknown, thereby potentially improving patient outcomes.
“Validation is a challenge because there is no ground truth. Existing methods failed to identify the origin. But the evidence we looked at showed us that the model is on the right track.”
– Alexander Gusev, PhD, Dana-Farber Researcher and Senior Author