Integrative Data Mining Pipeline for Identification of a Protein- Based Prognostic Signature in Lung Squamous Cell Carcinoma

Abstract The purpose of this study is to use an integrated data mining approach, in which multi-omics, clinical information, and image information are considered together, and to develop a new prognosis prediction model for Lung Squamous Cell Carcinoma (LUSC). We analyzed Reverse Phase Protein Array (RPPA) data of LUSC samples (n = 328) from The Cancer Genome Atlas cohort (TCGA). Univariate Cox regression analysis and the least absolute shrinkage and selection operator (LASSO) regression analysis followed by multivariate Cox analysis were performed to identify key protein candidates and constructed a robust multiprotein prognostic model on the training set. The optimal cut-off value was obtained by the receiver operating characteristic (ROC) curve, which was employed to divide patients into a high- and a low-risk group. The model was evaluated using multiple statistical methods, including principal components analysis (PCA), Kaplan-Meier survival analysis, independent prognostic analysis, ROC analysis, and immunohistochemistry (IHC) staining. The co-expression analysis and bioinformatics enrichment analysis of gene function was adapted to evaluate the prognostic effect and biological pathways of the model. Four-protein (Histone-H3, HSP27_pS82, CHK2, and PAXILLIN) prognostic signature was able to stratify patients into high- and low-risk groups with statistical significance. The signature estimates poor overall survival for high-risk patients in both training and testing sets. Histone-H3, HSP27_pS82, and CHK2 were found to be protective, while PAXILLIN was associated with poor prognosis. Univariate and multivariate Cox regression analysis showed that the risk model was an independent risk factor for overall survival (univariate: HR = 3.558, 95%CI = 2.451–5.169, p < 0.001, multivariate: HR = 2.515, 95%CI = 1.750–3.615, p < 0.001). The area under the curve (AUC) of the risk scores was 0.742. The correlation heatmap provided a landscape for 455 proteins. The gene set enrichment analysis (GSEA) results revealed that adhesion molecular and cancer pathways were enriched in the high-risk group and the cytochrome P450 pathway was enriched in the low-risk groups. Our finding discovered a set of novel 4-related prognostic signatures could serve as a sensitive independent prognostic factor for individualized survival predictions.