Translation of Gene Expression Data Into Personalized Treatment in Cervical Cancer: Machine Learning Approach

The majority of cervical cancers have been linked to the infection by human papillomavirus (HPV). There is a need to identify genes which play a role in the final manifestation of cervical cancers following HPV infection.

To identify a number of genetic markers associated with cervical cancer that may aid in the disease's diagnosis or prognosis using machine learning methods.

To do this, we will assess numerous gene expression profiles with integrative machine learning approaches such as random forest (RF) and support vector machine-based recursive feature elimination (SVMRFE). The conceptual analysis consists of following steps: (i) gene expression analysis and (ii) machine learning analysis for predicting genes.

The selected datasets were GSE75132 and GSE39001 for this study. Accuracy and cross validation were carried for both SVM-RFE and RF model for the gene identification purpose. R Bioconductor packages “GEOquery,” “limma,” and “umap” were utilized. The selected genes of machine learning methods were combined. The SVM model was the best for predicting the gene expression microarray profile based on the accuracy this study was able to get.

The SVM model indicated that genes might be used as biomarkers to identify biological processes. The identified genes were considered as potential gene signatures in cervical cancer detection, and their interactions were studied.