Predicting CNS penetration of precision medicine therapies in oncology: A comparison of the CNS TAP tool and the BOILED-Egg computational model.

Abstract

2062 Background: The number of available targeted therapies has vastly expanded in recent years; however, only a selection of such therapies achieve adequate central nervous system (CNS) penetration. The CNS Targeted Agent Prediction (CNS TAP) Tool was developed to aid in the selection of precision medicine therapies based on preclinical data, clinical data, patient-specific genomic data, and importantly, optimal blood brain barrier penetration. Using a standardized murine pharmacokinetic (PK) model, we sought to determine the concordance of CNS penetration prediction by the CNS TAP Tool compared to the SwissADME BOILED-Egg computational model. Methods: Eleven drugs were chosen based on the most commonly altered and targeted pathways by our CNS tumor board. Mice were injected individually with the compounds of interest and sacrificed at four different time points to quantify drug concentration via LC-MS in brain tissue and blood samples for PK analyses (1,2,4,7 hours). The percent brain penetration (AUCbrain/plasma) was compared to published human phase I PK data and pathway IC-50 to determine if agents were adequately CNS penetrant. These results were compared against the CNS TAP Tool and BOILED-Egg prediction of CNS penetrance. Results: All 11 drugs chosen (Table) demonstrated adequate CNS penetration, suggesting likely CNS efficacy. The CNS TAP Tool correctly predicted 10/11 agents (91% concordance), the lone exception being panobinostat. In contrast, the BOILED-Egg model only successfully predicted 3/11 agents (27%). When restricting the analysis to drugs with >50% CNS penetration (ponatinib, panobinostat, and ONC-201), the BOILED-Egg model successfully predicted CNS penetration of all agents. Conclusions: The clinician-curated CNS TAP Tool more accurately predicted CNS penetration of precision medicine therapies than the BOILED-Egg computational model, which performed best for drugs that achieve extremely high CNS penetration. The resulting PK and CNS penetration data for all agents was utilized to update and refine the CNS TAP tool for future patient use. [Table: see text]