For a small percentage of cancer patients, doctors are unable to find out where their cancer originated. This makes it rather more difficult to decide on a treatment for those patients, because many cancer drugs are typically developed for specific cancer types.

A recent approach developed by researchers at MIT and Dana-Farber Cancer Institute may make it easier to discover the sites of origin for those enigmatic cancers. Using machine learning, the researchers created a computational model that may analyze the sequence of about 400 genes and use that information to predict where a given tumor originated within the body.

Using this model, the researchers showed that they might accurately classify not less than 40 percent of tumors of unknown origin with high confidence, in a dataset of about 900 patients. This approach enabled a 2.2-fold increase within the variety of patients who might have been eligible for a genomically guided, targeted treatment, based on where their cancer originated.

“That was an important finding in our paper, that this model might be potentially used to help treatment decisions, guiding doctors toward personalized treatments for patients with cancers of unknown primary origin,” says Intae Moon, an MIT graduate student in electrical engineering and computer science who’s the lead creator of the brand new study.

Alexander Gusev, an associate professor of medication at Harvard Medical School and Dana-Farber Cancer Institute, is the senior creator of the paper, which appears today in .

Mysterious origins

In 3 to five percent of cancer patients, particularly in cases where tumors have metastasized throughout the body, oncologists don’t have a simple approach to determine where the cancer originated. These tumors are classified as cancers of unknown primary (CUP).

This lack of information often prevents doctors from with the ability to give patients “precision” drugs, that are typically approved for specific cancer types where they’re known to work. These targeted treatments are inclined to be simpler and have fewer unwanted side effects than treatments which are used for a broad spectrum of cancers, that are commonly prescribed to CUP patients.

“A sizeable number of people develop these cancers of unknown primary every yr, and since most therapies are approved in a site-specific way, where you’ve to know the first site to deploy them, they’ve very limited treatment options,” Gusev says.

Moon, an affiliate of the Computer Science and Artificial Intelligence Laboratory who’s co-advised by Gusev, decided to investigate genetic data that’s routinely collected at Dana-Farber to see if it might be used to predict cancer type. The information consist of genetic sequences for about 400 genes which are often mutated in cancer. The researchers trained a machine-learning model on data from nearly 30,000 patients who had been diagnosed with one in all 22 known cancer types. That set of knowledge included patients from Memorial Sloan Kettering Cancer Center and Vanderbilt-Ingram Cancer Center, in addition to Dana-Farber.

The researchers then tested the resulting model on about 7,000 tumors that it hadn’t seen before, but whose site of origin was known. The model, which the researchers named OncoNPC, was in a position to predict their origins with about 80 percent accuracy. For tumors with high-confidence predictions, which constituted about 65 percent of the full, its accuracy rose to roughly 95 percent.

After those encouraging results, the researchers used the model to investigate a set of about 900 tumors from patients with CUP, which were all from Dana-Farber. They found that for 40 percent of those tumors, the model was in a position to make high-confidence predictions.

The researchers then compared the model’s predictions with an evaluation of the germline, or inherited, mutations in a subset of tumors with available data, which may reveal whether the patients have a genetic predisposition to develop a selected variety of cancer. The researchers found that the model’s predictions were rather more more likely to match the variety of cancer most strongly predicted by the germline mutations than another variety of cancer.

Guiding drug decisions

To further validate the model’s predictions, the researchers compared data on the CUP patients’ survival time with the standard prognosis for the variety of cancer that the model predicted. They found that CUP patients who were predicted to have cancer with a poor prognosis, resembling pancreatic cancer, showed correspondingly shorter survival times. Meanwhile, CUP patients who were predicted to have cancers that typically have higher prognoses, resembling neuroendocrine tumors, had longer survival times.

One other indication that the model’s predictions might be useful got here from the sorts of treatments that CUP patients analyzed within the study had received. About 10 percent of those patients had received a targeted treatment, based on their oncologists’ best guess about where their cancer had originated. Amongst those patients, those that received a treatment consistent with the variety of cancer that the model predicted for them fared higher than patients who received a treatment typically given for a special variety of cancer than what the model predicted for them.

Using this model, the researchers also identified an extra 15 percent of patients (2.2-fold increase) who could have received an existing targeted treatment, if their cancer type had been known. As a substitute, those patients ended up receiving more general chemotherapy drugs.

“That potentially makes these findings more clinically actionable because we’re not requiring a recent drug to be approved. What we’re saying is that this population can now be eligible for precision treatments that exist already,” Gusev says.

The researchers now hope to expand their model to incorporate other sorts of data, resembling pathology images and radiology images, to supply a more comprehensive prediction using multiple data modalities. This is able to also provide the model with a comprehensive perspective of tumors, enabling it to predict not only the variety of tumor and patient final result, but potentially even the optimal treatment.

The research was funded by the National Institutes of Health, the Louis B. Mayer Foundation, the Doris Duke Charitable Foundation, the Phi Beta Psi Sorority, and the Emerson Collective.