Genomenon Creates Automated, Evidence-Based Cancer Gene Panel
Genomenon, a University of Michigan spinout developing analytics and data visualization software for the genomics industry, last week announced it has created an evidence-based cancer gene panel using automated machine learning techniques.
Gene panels are used by clinicians to analyze mutations in patient samples. Genomenon’s panel was produced using its software tool, the Mastermind Genomic Search Engine. Mastermind mines millions of medical journal articles to find associations between genetic mutations and disease, with the goal of significantly reducing the time doctors spend combing through medical data and simplifying the process of picking which genes to sequence. Genomenon’s gene panel for a range of blood cancers is the first such panel to be published using this approach, the company says.
The traditional method of selecting genes for sequencing involves a committee decision, whether from an academic or commercial group, to design a panel for a given suspected disease. The team then decides what genes should comprise the panel by doing manual searches of the literature, a process susceptible to human error and likely not as comprehensive as it could be, says Mark Kiel, Genomenon’s founder and chief scientific officer. In all, it can take months and, even then, may miss the mark.
“It’s a very laborious process and prone to adding and missing things,” Kiel explains. “We’re systemizing that approach and putting the value on evidence. Being able to do this repeatedly for any disease in an automated way will hopefully compel a bit more rigor to the process overall.”
Commercially available panels focus on a select set of genes that have known or suspected associations with the disease being studied.
“The commercial panels can be very different [from one another] even when it’s the same disease,” Kiel says. “We’ve taken an evidence-based approach to designing panels. We went to the source—medical literature—and asked a question: ‘What is the evidence from the literature that supports the idea that this gene is important to this disease?’”
Mastermind’s software uses algorithms to sift through the medical literature from publicly available sources, such as PubMed, and determines the relationship between specific diseases and gene mutations. Mastermind then organizes the data into clinical categories prioritized by the strength of those relationships, and then a specialist can review the list of gene candidates Mastermind comes up with for final approval.
Kiel says Mastermind’s ability to design panels for sequencing is the “beginning of the pipeline.” The software also asks if the mutation in question has ever been seen before, and, if so, when and how many times in order to give doctors a sense for how common the disease might be. Finally, Mastermind can suggest how doctors can clinically address the disease by analyzing gene-specific therapeutic, prognostic, and diagnostic information in the literature.
Genomenon presented its new blood cancer panel at the annual Association for Molecular Pathologists meeting last week. Earlier this year, the 15-person company received a phase II SBIR grant from the federal government worth $1.5 million to accelerate the process of getting the word out. Genomenon will be looking for additional outside investment in 2018, and the company is also seeking to publish its data in a top-tier peer-reviewed journal, Kiel says.
“Having access to all this data and knowing what to do with it opens all kinds of exciting possibilities for the future,” he adds.