Investigation of the clinical value of artificial intelligence-derived prognostic signature in cervical cancer based on machine learning algorithms

Women around the world are troubled by life-threatening cervical cancer. It is urgent to identify a biomarker to improve the prognosis of cervical cancer patients. Based on gene expression profiles and single-cell sequencing data obtained from public databases, we performed dimensionality reduction and clustering analyses, Scissor analysis, WGCNA, and machine-learning modeling using 10 base algorithms and their 101 individual or combined strategies. We finally screened 24 consensus prognostic genes to develop a novel model artificial intelligence-derived prognostic signature (AIDPS), based on C-index which was detected in six validation datasets (TCGA_Test, TCGA_Entire, CGCI-HTMCP-CC, GSE39001, GSE44001, and GSE52903). AIDPS demonstrated modest but consistent prognostic performance across multiple independent cervical cancer cohorts, with an average C-index of 0.665.The accuracy of AIDPS in predicting CESC was significantly better than that of other clinical characteristics including age, pathological TNM stages, and grade. In conclusion, our study developed a consensus model AIDPS, an effective strategy to further guide the clinical management and individualized treatment of cervical cancer.