Abstract B065: Stage-aware CRISPR design for brain metastasis interception: Multi-modal validation of BBB extravasation and CNS colonization targets

Abstract Background: Brain metastases occur in 20-40% of patients with breast, lung, and melanoma cancers and represent a leading cause of cancer mortality. Current CRISPR design tools prioritize on-target efficiency alone, ignoring stage-specific vulnerabilities along metastatic cascades. We present Interception, a stage-aware CRISPR design framework that can be applied to brain colonization by integrating multi-modal genomic signals and foundation models to target BBB extravasation and CNS microenvironment adaptation. Methods: We implemented a modular pipeline that (i) computes Functionality (Evo2 variant predictions), Essentiality (DepMap scores), Chromatin accessibility (framework for CNS-specific enhancers), and Regulatory impact; (ii) selects mission-specific targets using weighted Target-Lock scoring; (iii) generates PAM-aware guide candidates; (iv) scores efficacy via Evo2 delta predictions; (v) quantifies genome-wide safety via minimap2/BLAST with exponential decay mapping. Candidates ranked by composite Assassin score: 0.40×efficacy + 0.30×safety + 0.30×mission fit. To demonstrate applicability to brain metastasis, we identified candidate genes for BBB extravasation (VEGF-C, MMP9) and CNS adaptation (L1CAM, STAT3, CTGF) based on published brain metastasis literature. Results: We validated the Target-Lock scoring framework using 38 metastasis-associated genes across 8 cascade steps (304 data points). Per-step AUROC was 0.976 ± 0.035, AUPRC 0.948 ± 0.064, Precision@3 = 1.000 (1000-bootstrap CIs). All 8 steps showed significant enrichment (Fisher's exact p < 0.05, 6/8 with p < 0.001), with large effect sizes (Cohen's d > 2.0). Guide RNA designs (n=20) achieved mean efficacy 0.548 ± 0.119, safety 0.771 ± 0.210, Assassin score 0.517 ± 0.114. Structural validation of 15 guide:DNA complexes via AlphaFold 3 achieved 100% pass rate (pLDDT 65.6 ± 1.8, iPTM 0.36 ± 0.01; acceptance: pLDDT ≥50, iPTM ≥0.30). When applied to brain metastasis candidate genes, the framework prioritizes targets implicated in BBB permeability (VEGF-C), neuronal adhesion (L1CAM), and astrocyte-mediated survival (STAT3) based on Target-Lock scoring. Conclusions: Interception delivers a reproducible, mission-aware CRISPR design framework validated on metastatic cascade genes (AUROC 0.976, 100% structural pass rate) and applicable to brain metastasis target selection. The framework integrates multi-modal signals, genome-wide safety validation, and structural confirmation. Application to brain colonization identifies candidate targets for BBB extravasation and CNS microenvironment adaptation. Future work includes experimental validation in brain-tropic cell lines (MDA-MB-231-BR, PC9-BrM3), integration of brain-specific DepMap essentiality, and in vivo intracranial injection models to measure metastatic burden reduction. Citation Format: Fahad Kiani, Sabreen Abeed Allah. Stage-aware CRISPR design for brain metastasis interception: Multi-modal validation of BBB extravasation and CNS colonization targets [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(6_Suppl):Abstract nr B065.