A Prognostic Risk Score Based on Integration of 9-Gene Expression and 5-Microbial Abundance Improves the Power to Predict Overall Survival of Patients with Cervical Squamous Carcinoma

Background: Prognosis of cervical squamous cancer (CESC) patients varies widely. Therefore, it is urgent to identify a reliable prognostic biomarker to distinguish patients with poor prognosis. Methods: Transcriptionally altered genes in CESC were identified by GEO2R in NCBI GEO database. Kaplan-Meier and Cox regression analysis was used to assess association of gene expression, microbial abundance, and scores with overall survival (OS). The multivariate Cox regression with/without resampling method was used to identify a gene or microbe signature associated with OS. Prognostic scoring systems were developed. Findings: We found 308 genes both transcriptionally altered in CESC and significantly correlated to OS. Resampling 100 times, we identified a 9-gene prognostic signature (SLC25A28, RFC5, RNASEH2A, SLC39A14, APOBEC3B, MS4A7A, FAM20C, ERG and TFRC) in TCGA-CESC dataset. A 9-gene expression-based prognostic score (GPS) was created and externally validated to predict OS. Moreover, our GPS performed significantly better than five published gene prognostic signatures. Similarly, we identified a 5-microbe prognostic signature (Dermabacter, Sulfurovum, Nostoc, Thermacetogenium and Alphavirus) using TCGA-CESC microbiome. A 5-microbial abundance-based prognostic score (MPS) was created to predict OS. When integrated the 9-gene signature with 5-microbe signature, the gene and microbe prognostic score (GMPS) showed much better performance for predicting OS than GPS and MPS. Interpretation: Our study constructed a GPS, MPS and GMPS for robustly recognizing patients with poor prognosis, which could guide individualized treatment in the future. Funding: National Natural Science Foundation of China, Zhongnan Hospital of Wuhan University Science, Technology and Innovation Seed Fund. Declaration of Interest: The authors declare no conflict of interest.