Fig. 7: Cost savings analysis.

A Percentage of molecular test reduction vs. biomarker prevalence using the high-sensitivity model presented in Fig. 4A (88.7% sensitivity, 31.8% specificity). In this scenario, patients predicted biomarker negative would be de-prioritized from molecular testing (i.e., rule-out patients in a clinical study with pre-defined sites like ANNAR). B Percentage of FGFR+ detected patients vs. percentage of patients receiving molecular test using different operating points of our FGFR model compared to without AI. Each curve represents enhancement for a given biomarker prevalence. Red and blue dotted lines represent the “Low” and “High” operating points of the simulated 3-tier model presented in Fig. 4D. C Prevalence enrichment using the simulated 3-tier model. Note that “High” vs. Baseline shows enrichment in the”High” group compared to baseline prevalence, whereas “High” vs “Low” shows prevalence enrichment of “High” group vs. “Low” group.