Researchers at Dana-Farber Cancer Institute have developed an artificial intelligence (AI) program that may help trace cancers of unknown primary origin back to their sources.
Cancers of unknown primary origin make up about 3% to 5% cases of metastatic disease; they are cancers that have spread from a separate and unknown place in the body that cannot be determined through conventional imaging scans, biopsies, or pathology reports.
The AI program, OncoNPC, uses sequencing data from tumor DNA to determine where the tumor came from. The machine-learning classifier was developed with targeted next-generation sequencing (NGS) data from 36,445 tumors across 22 cancer types and medical records from the patients.
In results from a study published recently in Nature, OncoNPC was able to accurately predict the origin of about 80% of the tumors tested, including metastatic disease. Furthermore, OncoNPC correctly identified about 95% of the findings that the model considered high-confidence predictions—65% of the samples used for validation.
Cancers of unknown primary origin are particularly hard to treat, as most therapies are targeted at specific types of tumors; the researchers aim to mitigate that with OncoNPC. Determining the origin and type of tumors gives patients better options for treatment, leading to better outcomes.
“This patient group has dismal outcomes,” said researcher and senior author Alexander Gusev, PhD, in a statement. ““We see the OncoNPC prediction as a nudge, a way to provide a possible explanation for the cancer that helps point to appropriate treatment, including precision medicine.”
The researchers plan to refine OncoNPC’s performance through the addition of other types of information, such as pathology reports, as well as exploring how the model may complement other diagnostic techniques in the community cancer treatment setting.
