Multi-omic integration identifies mechanisms of Broad Drug Resistance and opportunities for therapeutic reprogramming cancer cells

Broad drug resistance in cancer arises through diverse transcriptional, metabolic, and genetic adaptations, yet the shared molecular programs that sustain cross-resistant phenotypes remain incompletely defined. This study integrates PRISM drug-response data with transcriptomic, metabolomic, and mutational profiles to characterize the molecular features associated with broad drug resistance and to identify compounds capable of reversing resistance-associated gene signatures. Resistant cell lines exhibited coordinated activation of extracellular matrix remodeling, stress-adaptation pathways, and survival signaling, with NFE2L2 emerging as a central regulatory hub linking upstream mutations to oxidative-stress transcriptional programs. Multi-omic integration further revealed metabolic reprogramming as a conserved hallmark of resistance, and analyses of clinical cohorts demonstrated that resistance-associated alterations were associated with reduced progression-free survival. Computational perturbagen screening nominated compounds predicted to counteract resistance-associated transcriptional signatures. Experimental validation confirmed that rosiglitazone suppressed NFE2L2-associated gene expression programs and restored chemotherapy sensitivity in resistant models, supporting a scalable framework for rational phenotypic reprogramming. This GEO submission provides raw RNA-seq data generated from compound-treated OE19 cells used to experimentally validate candidate re-sensitization strategies. Overall design: OE19 esophageal adenocarcinoma cells were treated with 1 µM rosiglitazone, halcinonide, methylprednisolone or 0.1% DMSO (control) , each in biological duplicate, alongside two untreated control samples. Total RNA was extracted using the Qiagen RNeasy Mini Kit according to the manufacturer's instructions. RNA-seq libraries were prepared by the University of Notre Dame Genomics Core and sequenced on an Illumina NextSeq 2000 platform (paired-end, PE50), generating ~30 million reads per sample. Raw FASTQ files were aligned to the GRCh38/hg38 reference genome using HISAT2, and gene-level read counts were produced using featureCounts. FASTQ files (R1 and R2 for each biological replicate) are provided as supplementary data to support reproducibility and downstream analysis.