Integrating AI and Transcriptomics to Identify Biomarkers of Metastatic Cervical Cancer in Patient Serum

Abstract

Metastatic cervical cancer is regulated by soluble immune-modulating factors that are difficult to detect in patient serum. Standard transcriptomic analysis provides data, but the integration of machine learning (ML) and deep learning (DL) can also aid in the identification of a serum-induced transcriptional signature of metastasis. In this research, ML/DL methods will be used to discriminate between local and metastatic cervical cancer using gene expression profiles. NCBI GEO gene expression data were analyzed through differential expression analysis and machine learning-based feature selection techniques, such as LASSO regression, Random Forest, and Boruta algorithm. Features selected were employed to train ML classifiers, such as Support Vector Machines (SVM), XGBoost, and Artificial Neural Networks (ANNs).Deep learning approaches, for example, autoencoders and LSTM networks, were considered. Functional enrichment and pathway analysis were conducted using IPA and KEGG to find key biological regulators, such as IL-10, immunoglobulins, and miRNAs (miR-23a-3p, miR-944). ML models correctly classified metastatic vs local cervical cancer with high accuracy, and XGBoost was the best classifier (AUC > 0.90). Downregulation of immune surveillance processes, such as phagocytosis and hematopoietic cell accumulation, was identified as major metastatic markers by analysis. Feature extraction using deep learning further enhanced classification to detect hidden transcriptomic patterns related to metastasis. This study demonstrates that ML/DL approaches can differentiate metastatic cervical cancer by analyzing serum-induced transcriptional signatures efficiently. The use of AI-based models in combination with traditional bioinformatics can assist in biomarker identification and improve predictive diagnostics for cervical cancer metastatic progression.

Keywords: Boruta algorithm, Cervical cancer, LASSO regression, Machine learning.

View PDF