Integrative miRNA-Gene expression analysis enables refinement of associated biology and prediction of response to cetuximab in head and neck squamous cell cancer

Abstract

This paper documents the process by which we, through gene and miRNA expression profiling of the same samples of head and neck squamous cell carcinomas (HNSCC) and an integrative miRNA-mRNA expression analysis, were able to identify candidate biomarkers of progression-free survival (PFS) in patients treated with cetuximab-based approaches. Through sparse partial least square-discriminant analysis (sPLS-DA) and supervised analysis, 36 miRNAs were identified in two components that clearly separated long-and short-PFS patients. Gene set enrichment analysis identified a significant correlation between the miRNA first-component and EGFR signaling, keratinocyte differentiation, and p53. Another significant correlation was identified between the second component and RAS, NOTCH, immune/inflammatory response, epithelial-mesenchymal transition (EMT), and angiogenesis pathways. Regularized canonical correlation analysis of sPLS-DA miRNA and gene data combined with the MAGIA2 web-tool highlighted 16 miRNAs and 84 genes that were interconnected in a total of 245 interactions. After feature selection by a smoothed t-statistic support vector machine, we identified three miRNAs and five genes in the miRNA-gene network whose expression result was the most relevant in predicting PFS (Area Under the Curve, AUC = 0.992). Overall, using a well-defined clinical setting and up-to-date bioinformatics tools, we are able to give the proof of principle that an integrative miRNA-mRNA expression could greatly contribute to the refinement of the biology behind a predictive model.