Cell line screening of AI-generated drug repurposing predictions for NF2-related schwannomas

Abstract

Non-NF2 schwannomatosis (SWN) is a rare syndrome that causes nerve sheath tumors with complex genetic variations and poorly understood modifiers. With no approved pharmacologic interventions, there is an urgent need for treatments that prevent or reduce tumor growth, or that ameliorate schwannoma-associated pain. Many NF transcriptomic datasets lack patient-matched non-tumor tissue for comparison. We piloted a methodology to compare non-patient-matched expression data and performed differential expression analysis between Synodos SWN schwannoma samples and GTEx normal tibial nerve samples. We performed functional enrichment analysis of differentially expressed genes to identify potential drug targets. To complement these data-driven predictions, we employed TxGNN, an AI model for rare disease drug repurposing, to identify additional drug candidates. We identified 34 significantly upregulated genes in schwannomas from patients with pain compared to tibial nerve samples that are not upregulated in schwannomas from patients without pain. Drug target enrichment analysis of pain-associated genes identified pantiumumab, cetuximab, and pembrolizumab. Next, we used TxGNN to identify 198 drug candidates, which includes the three drugs identified in differential expression analyses. We used large language models to filter this list to 94 available small molecules that have not been previously tested in schwannomas and are not cytotoxic chemotherapies. We are collaborating with investigators at UCF to validate these predictions with high-throughput screens of candidate compounds in human merlin-deficient schwannoma cells using automated confluence monitoring. Top candidates will undergo mechanistic validation in primary patient-derived schwannoma cells coupled with transcriptomic analysis to map affected pathways and establish criteria for future drug prioritization.