Network analysis and machine learning approach identifies HSP90-associated synthetic lethal targets and drug responses in breast cancer

Heat shock protein 90 (HSP90) is an ATP-dependent molecular chaperone that stabilizes numerous oncogenic proteins involved in proliferation, survival, DNA repair, and therapeutic resistance. Despite its central role in breast cancer (BRCA), monotherapy targeting HSP90 has shown limited clinical success, highlighting the need to identify therapeutically actionable HSP90-associated vulnerabilities. Here, we integrated transcriptomic, network, pharmacogenomic, single-cell, and CRISPR dependency datasets. The major HSP90 isoforms, HSP90AA1 and HSP90AB1, were significantly overexpressed in BRCA and independently associated with poor survival. Immune and single-cell analyses linked high HSP90 expression to an immune-cold tumor microenvironment and tumor cell programs related to proliferation, DNA damage response, invasion, and metastasis. Network analysis identified synthetic lethal/viable (SL/SV) interactions involving key regulators, including PAK1, MAP2K1, MAPK3, NFKB1, TNF, CALN1, and HACD3, highlighting interconnected therapeutic vulnerabilities. Pharmacogenomic integration with GDSC and six machine-learning algorithms identified 26 candidate compounds predicted to be effective across 51 breast cancer cell lines. CRISPR dependency analyses demonstrated negative gene effect scores for HSP90 and several SL partners, supporting their essential roles in tumor cell viability. These findings establish HSP90 as a central regulatory hub and provide a rationale for mechanism-based combination therapies and computational prioritization of precision oncology targets for experimental validation.