Open source machine learning tool helps clinicians choose chemo drugs
Figuring out which chemotherapy drug to use when the first-line chemotherapy drug fails is a complicated decision for clinicians. EurekAlert! reports that Georgia Institute of Technology researchers have built an open source machine learning tool that can help them make data informed decisions.
According to EurekAlert!, the tool uses machine learning to analyze RNA expression linked to information about patient outcomes with specific drugs, thereby generating suggestions for chemotherapy drug with the highest likelihood of success.
The system predicted the chemotherapy drug that had provided the best outcome with 80 percent accuracy, per a study involving RNA analysis data from 152 patient records. Researchers believe that adding more patient records as well as family history and demographics could boost the system’s accuracy.
"By looking at RNA expression in tumors, we believe we can predict with high accuracy which patients are likely to respond to a particular drug. This information could be used, along with other factors, to support the decisions clinicians must make regarding chemotherapy treatment."
"By looking at RNA expression in tumors, we believe we can predict with high accuracy which patients are likely to respond to a particular drug," said John McDonald, a professor in the Georgia Tech School of Biological Sciences and director of its Integrated Cancer Research Center. "This information could be used, along with other factors, to support the decisions clinicians must make regarding chemotherapy treatment."
The Atlanta-based Ovarian Cancer Institute, the Georgia Research Alliance, and a National Institutes of Health fellowship contributed funding towards this research, published last week in the journal Scientific Reports, which adds to growing literature on precision medicine for cancer treatment.
Though the research started with a data set focused on ovarian cancer, but ultimately the research team expanded the data set to encompass other cancers such as lung, breast, liver and pancreatic cancers.
"Our model is predicting based on the drug and looking across all the patients who were treated with that drug regardless of cancer type," McDonald explained.
McDonald also believes the declining cost of RNA analysis (projected to drop below the cost of a mammogram) could give it an advantage over DNA sequencing.
His team hopes that hospitals and cancer centers will leverage this tool when it becomes available as open source software, especially since its accuracy will improve with more patient data.
"We are trying to create a different paradigm for cancer therapy using the kind of open source strategy used in internet technology," McDonald said.