Genialis krasID analyzes foundational aspects of KRAS biology using RNA-sequencing and a machine learning algorithm comprising numerous biologic signatures. The classifier predicted preclinical drug response with an accuracy of 0.94 Area Under the Receiver Operator Curve (AUROC) and correctly selected real-world patient responders with an AUROC of 0.80. Kaplan-Meier analysis of the real-world sotorasib patient data considered time on treatment as an available proxy for progression-free survival. This analysis showed that patients classified as “krasID-high” had median time on treatment of almost one year (338 days), compared with those classified as “krasID-low” with median time on treatment of only 158 days (hazard ratio [HR] = 0.35). Taken together, these early results demonstrate that the Genialis krasID classifier can predict response and stratify durable benefit in experimental and clinical settings.

Published at AACR 2024.