Figure 4 from GPSai: A Clinically Validated AI Tool for Tissue of Origin Prediction during Routine Tumor Profiling

Clinical relevance of GPSai. A–D, Quantification of diagnosis changes and level 1 targeted therapy recommendation changes prompted by GPSai. A, Cases with “critical value discrepancies” and diagnosis changes supported by orthogonal evidence out of all profiled cases between March and October, 2024. B, Cases in which the GPSai call was confirmed by select orthogonal methods are shown. C, Percentage of cases with changes in level 1 targeted therapy recommendations among all cases with diagnosis change prompted by GPSai. Although some cases were eligible/ineligible for more than one drug, no more than one drug per case was included in this analysis. D, Breakdown of the technologies used to identify the biomarker for level 1 targeted therapy treatment associations. If drug eligibility could be driven by WES/copy-number alterations (CNA) or WTS, that drug was chosen first, followed by IHC-driven drugs. If no biomarker-driven drug association changes were present, diagnosis-only changes were recorded. E, Physicians who received a GPSai “critical value discrepancy” on the clinical report for their patient were surveyed to understand how the result affected patient care. There were 97 survey responses corresponding to 97 unique patients.