Cancer Data, Conveniently

Computer science Assistant Professor Jacob Luber is creating a database that will help researchers and physicians map where cancer patients have similar traits and improve and expand treatments based on that data. The result will contain every publicly available cancer dataset from the National Cancer Institute.

Such datasets are currently too large for physicians to access easily. To overcome this, Dr. Luber is developing an algorithm that compresses the bioinformatic data, which would allow doctors to input new information and pull up the larger index through a computer at a patient's bedside or in an exam room. Additionally, the imaging index will be used to further drug-discovery efforts by providing a better understanding of how chemicals derived from the microbiome affect an immune system's ability to fight cancer.

"With deep learning, we can take a high-dimensional image with proteomic data overlaid on top and compress it, then use it to find similarities between patients,"" says Luber. "Once those similarities are identified, physicians could mine electronic health records to see what treatments were most effective and adjust treatment regimens accordingly."