Biomarker platform is among the first validated applications of Genialis’ machine learning approach to help predict patient responses to targeted oncology therapies
Genialis, a leader in applied data science for the development of precision medicines, today announced its contributions to a biomarker platform and panel that demonstrated prognostic capabilities for recurrence-free and overall survival in colorectal cancer (CRC). In a poster presentation (#348) at the virtual AACR Annual Meeting 2021: Discovery Science Driving Clinical Breakthroughs, Genialis helped support the evaluation of OncXerna Therapeutic’s Xerna™ platform and TME Panel, an RNA-based pan-tumor biomarker, in two gene expression datasets—one from 566 patients and the other from 93 patients. This novel diagnostic and prognostic platform uses a proprietary set of biomarker genes, RNA-based gene expression data, and machine learning to classify patients based on the dominant biologies of the tumor microenvironment.
“Working with the team at Genialis has been incredibly productive as we apply their expertise in data curation, feature analysis and machine learning to the development of our pan-tumor Xerna platform and TME panel,” said Laura Benjamin, Ph.D., president and chief executive officer at OncXerna Therapeutics. “We are excited to continue to work with Genialis as we advance our platform and multiple clinical programs to expand the use of precision medicine to more people with cancer.”
An estimated 96 to 97 percent of cancer drugs that enter clinical trials fail at some phase of development, representing a huge loss of capital and opportunity cost, not to mention the toll on patient trial enrollees in grave need of effective treatment. However, it has also been shown that clinical trials in oncology are up to 10 times more likely to meet their endpoints if guided by a biomarker. One key application area of Genialis technologies has been to identify and remove biases that might arise from sources such as demographics, technical assays, and disease/tissue of origin.
“Most biomarkers don’t translate to help real patients, in part because of bias introduced through feature selection and training data. This is a hard but fundamental problem for which we’re developing solutions—how to ensure predictive models are widely applicable by addressing sources of bias head on,” said Rafael Rosengarten, Ph.D., chief executive officer of Genialis. “The goal of precision medicine is to help all patients get the best medicine for their unique disease biology. Applying a biomarker to predict possible therapy outcomes in many disease areas gives us more opportunities to make a difference.”
Genialis deploys a complete technology stack for biomarker discovery and development, including its Expressions software platform and proprietary technologies for data processing and machine-learning enablement. In addition, the company specializes in interactive visualizations to support data mining and “unboxing” black box models to explain the underlying AI. The company recently announced a sponsored research agreement with BioLab, a data science research group at the University of Ljubljana, to advance its patient stratification toolkit. With its machine-learning guided approach to biomarker development, Genialis aims to make drug development more successful, and give patients in need a fighting chance.
