Integrating single-cell and spatial transcriptomics with computational perturbation identifies candidate prognostic genes associated with mitochondrial function and succinylation-related annotations in ovarian cancer

Ovarian cancer (OC) is the deadliest gynecological malignancy worldwide. Whether mitochondrial succinylation—an important post-translational modification—plays any functional role in this disease remains largely unclear. To address this gap, we integrated TCGA and GEO transcriptomic data with single-cell RNA-seq and spatial transcriptomics, and applied in silico gene perturbation to characterize candidate prognostic genes identified from overlapping mitochondrial-function and succinylation-related annotations in OC. We then systematically characterized the cellular origins, spatial distributions, and potential biological functions of the identified genes. Four genes—ACSS3, SIRT3, SHMT2, and SIRT5—were identified as candidate prognostic genes. A risk score derived from their expression effectively stratified OC patients into high- and low-risk groups with markedly different survival outcomes. Single-cell analysis revealed that ACSS3 is mainly expressed in fibroblasts and epithelial cells; SIRT3 is more broadly distributed across epithelial, fibroblast, endothelial, and T/NK cell populations; SHMT2 is enriched in epithelial, endothelial, and fibroblast cells; and SIRT5 is predominantly found in epithelial, fibroblast, and endothelial cells. Spatial transcriptomics confirmed that these genes are enriched in specific tumor microenvironment regions. RT-qPCR on clinical OC specimens verified that ACSS3, SHMT2, and SIRT5 are overexpressed in tumors. However, all functional inferences regarding these genes were derived solely from in silico perturbation analyses and had not been validated by cellular knockdown/overexpression, pharmacological intervention, or animal model experiments; accordingly, their direct causal regulatory roles or therapeutic target value could not be established from the present data. The risk model may assist in patient stratification after further optimization and prospective validation.