On June 12, 2023 Genialis, a computational precision medicine company unraveling complex biology to find new ways to treat disease, reported results from an independent validation study of a KRAS biomarker signature for lung adenocarcinoma (Press release, Genialis, JUN 12, 2023, View Source [SID1234632687]). In this study, Genialis reimplemented a previously developed gene expression-based classifier and applied it to a new data set to evaluate the model’s reliability and applicability across demographically distinct clinical cohorts. While the original study’s results could be faithfully recapitulated, the classification scheme did not transfer to a new cohort with different demographics and treatment histories. These results demonstrate a need for a more robust or generalizable set of features and models. A new poster with the results was released at the 2023 Annual Congress of the European Association for Cancer Research (EACR23-0944).
Schedule your 30 min Free 1stOncology Demo!
Discover why more than 1,500 members use 1stOncology™ to excel in:
Early/Late Stage Pipeline Development - Target Scouting - Clinical Biomarkers - Indication Selection & Expansion - BD&L Contacts - Conference Reports - Combinatorial Drug Settings - Companion Diagnostics - Drug Repositioning - First-in-class Analysis - Competitive Analysis - Deals & Licensing
Schedule Your 30 min Free Demo!
A search of PubMed returns more than one million entries for papers related to biomarkers, showing an exponential increase in the database since 1980. Interest in biomarkers is understandable since using biomarkers increases the odds of advancing an oncology clinical asset to the next trial phase by 5-12 fold. Further, even once approved, many of the best drugs achieve clinical benefit in only 30-40 percent of patients. Yet published biomarkers often prove to be overfitted to particular datasets and perform poorly on additional patient cohorts, particularly those with diverse genetic backgrounds. Today’s Genialis-EACR poster underscores the need to address biases in data collection and model training when developing novel biomarkers to ensure clinical utility.
"Most investigational cancer drugs, up to 97 percent, fail in clinical trials. And of those that succeed, many benefit only a fraction of patients," said Luka Ausec, Ph.D., Chief Discover Officer at Genialis. "Our ResponderID platform is designed specifically to enable the discovery, training, and validation of machine learning models as predictive biomarkers. We are developing a new generation of biomarkers that use gene expression data to represent complex disease biologies, and the work often starts with improving upon the existing state-of-the-art models to help meet the needs of more patients. The present study exemplifies how thorough study can expose weaknesses in the prior art."
The ResponderID platform is a technology suite for clinical and translational research, built from years of experience working with partners across the industry and advanced internal R&D. With a growing library of standard of care and state-of-the-art signatures built in, it provides clinical and translational researchers with a comprehensive molecular portrait of their patients, yielding the most informed decision-making possible.
To request a meeting at EACR, or for more information on ResponderID, please email [email protected] or visit www.genialis.com.