A novel prognostic signature of coagulation-related genes leveraged by machine learning algorithms for lung squamous cell carcinoma

Abstract

Coagulation-related genes (CRGs) have been demonstrated to be essential for the development of certain tumors; however, little is known about CRGs in lung squamous cell carcinoma (LUSC). In this study, we adopted CRGs to construct a coagulation-related gene prognostic signature (CRGPS) using machine learning algorithms. Using a set of 92 machine learning integrated algorithms, the CRGPS was determined to be the optimal prognostic signature (median C-index = 0.600) for predicting the prognosis of an LUSC patient. The CRGPS was not only superior to traditional clinical parameters (e.g., T stage, age, and gender) and its commutative genes but also outperformed 19 preexisting prognostic signatures for LUSC on predictive accuracy. The CRGPS score was positively correlated with poor prognoses in patients with LUSC (hazard ratio > 1, p < 0.05), indicating its suitability as a prognostic marker for this disease. The CRGPS was observed to be inversely correlated with the degree of infiltration of natural killer cells. For some tumors, patients with lower CRGPS scores are more likely to experience enhanced immunotherapy effects (area under the curve = 0.70), which implies that the CRGPS can potentially predict immunotherapy efficacy. A high CRGPS score is predictive of an LUSC patient being sensitive to several drugs. Collectively, these findings indicate that the CRGPS may be a reliable indicator of the prognoses of patients with LUSC and may be useful for the clinical management of such patients.

Keywords

• Cancer
• Prognosis
• Immunotherapy
• mRNA
• Protein