A team at Vanderbilt University Medical Center (VUMC) has developed and tested software that scans electronic health records in real time to monitor cancer patient survival (from time of diagnosis) according to which genes, if any, are found to carry mutations.
The software is called CUSTOM-SEQ, which stands for Continuously Updating System for Tracking Outcome by Mutation, to Support Evidence-based Querying. The software also tracks tumor type (anatomic site), cancer stage at diagnosis and smoking status.
CUSTOM-SEQ can be viewed as a fledgling version of a so-called learning system for cancer evaluation and treatment.
“Tracking tumor mutations and survival would be central to any learning system for cancer. But you might also be interested in things like cost, drug toxicities, progression of cancer and utilization of resources. Survival is probably the thing patients care about the most, so that’s why we gravitated to that,” said a member of the team that conceived the software, Jeremy Warner, M.D., M.S., assistant professor of Medicine and Biomedical Informatics.
Warner and colleagues recently described a pilot of CUSTOM-SEQ in the Journal of the American Medical Informatics Association. In the pilot, the software scanned records of 4,310 cancer patients who received tumor genotyping at VUMC between 2010 and 2015, with median follow-up of 17 months.
Two statistically significant patterns emerged. As expected, epidermal growth factor receptor (EGFR) mutations in lung cancer were associated with superior overall survival. In a novel finding, guanine nucleotide binding protein, q polypeptide (GNAQ) mutations in melanoma were associated with markedly inferior overall survival.
Warner qualifies these findings: “We’re algorithmically determining dates of diagnosis and death, those algorithms are validated but still have errors, and we freely admit that. The results aren’t intended to be a strong scientific finding. CUSTOM-SEQ is really meant to prompt a further investigation into potentially significant findings.
“The intent of the study was to describe this informatics approach. It certainly was interesting to have a new finding in regard to melanoma survival. But, for now, we’re using a fairly crude approach. We’re lumping all the mutations in a single gene together, when we know that different mutations on the same gene can pose contrasting signals. Also, much of the survival that we’re measuring is modified by treatment exposure, and tracking those exposures was out of scope for the initial pilot.”
Warner said it’s somewhat surprising that no other gene mutations had a statistically significant association with survival, adding that this may in part be due to the pilot’s population size, with many of the mutation-defined groups of patients having remained too small to exhibit distinct survival patterns.
Warner said the tool remains in development, and needs to be modified to handle information from the newer, more extensive genetic panels now in use in cancer. According to Warner, newer clinical assays typically include full sequencing of 200 to 500 genes, capturing not just point mutations, but also structural mutations such as copy number variations.
Warner was joined in the study by Lucy Wang, William Pao, M.D., Ph.D., Jeffrey Sosman, M.D., Ravi Atreya, Pam Carney, and Mia Levy, M.D., Ph.D.