Mathematical modeling to fight breast cancer
Virginia Tech researchers have developed a model that may eventually help explain and overcome drug resistance in breast cancer treatment.
Since every tumor is different, predicting the best treatment for a particular person’s cancer is not an easy task.
Cancers are traditionally treated with surgery, radiation, and chemotherapy, which kill any cells that divide rapidly, including cancer, hair follicle, bone marrow, and digestive tract cells. More recently, drugs like tamoxifen are being used to interfere with different molecules or functions that cancer cells need. This targeted therapy is used to prevent or postpone the cancer from growing or metastasizing to a different organ.
Sometimes one treatment is sufficient, and sometimes a mix of treatments is needed. Unfortunately, it often takes trial and error to determine which treatments are most effective for a particular patient. Patients must endure the treatment and its side effects before doctors can determine its effectiveness. This can take weeks.
Even when the treatment is initially successful, many cancers then become resistant to the treatment and the process of selecting a treatment must begin again.
Using a computer model to figure out which drugs are most likely to help promise to make selection of treatments a faster and more accurate process. Computers might be able to determine in seconds what could take months of trial-and-error.
Researchers working on this problem use machine learning techniques to look at genetic and tumor data from previous cancer patients. The computer examines the data, and groups them according to similar features (such as tumor size and composition, or which genes are most active). The computer system then fits current cancer patients into these categories. By knowing how previous patients in those categories responded to treatment, they can predict what treatments are likely to be effective.
Even with high-powered computers, grouping the patients is still a very hard problem to solve. There is no perfect prediction algorithm yet, but researchers are making steady progress and are already developing recommendations that can help patients in the near term.
The ultimate goal is precision medicine, where doctors take a sample of tumor, get it analyzed, and have the optimal treatment suggested for the precise tumor in any individual.