Multi-modal analysis of host transcriptomic signatures to decode etiology-specific marker genes in vaginal infections

Vaginal infections of bacterial, fungal, and viral origin often exhibit overlapping clinical manifestations, leading to diagnostic ambiguity and suboptimal treatment. Despite advances in microbiological testing, there remains a critical gap in host-based molecular markers that can reliably distinguish infection etiology. Addressing this limitation, we developed an integrative multi-modal transcriptomic framework to decode infection-specific host response signatures. Publicly available transcriptomic datasets were systematically analyzed using differential expression profiling and weighted gene co-expression network analysis (WGCNA) to capture infection-associated gene modules. These modules were further refined through protein–protein interaction network analysis and ModuLand-based core network extraction and functional enrichment analysis was performed to delineate key biological processes. To overcome challenges of high-dimensional data and multicollinearity, we employed an elastic net regression model integrating normalized expression values and gene set enrichment scores, enabling precise biomarker prioritization. Rigorous cross-validation ensured model robustness, and genes surpassing a 95% disease association probability threshold were selected. Multi-classifier machine learning also demonstrated strong discriminatory performance across the selected genes of three infection types. Through this multimodal analysis, we revealed distinct host signatures- bacterial infections were marked by mostly ribosome biogenesis genes (RRS1, CDC25A, LTV1, RSL1D1), fungal infections by antigen presentation genes (HLA-E, HLA-F), and viral infections by immune modulation and membrane remodeling genes (A2M, TRAIP). This study presents a systems-level framework integrating network biology and predictive modeling to identify etiology-specific host signatures, which, upon future validation, may enable a multi-gene panel for precise discrimination of bacterial, fungal, and viral vaginal infections.