Integrated machine learning constructed a circadian-rhythm-related model to assess clinical outcomes and therapeutic advantages in hepatocellular carcinoma.

Abstract

Background: Circadian rhythm (CR) coordinates a variety of internal biological processes with the external daily cycles of light and dark. However, the implications of CR-related regulator in hepatocellular carcinoma (HCC) are quite obscure. Here, we aimed to identify pivotal CR-related markers in HCC for predicting survival and treatment outcomes.

Methods: The prognostic value of CR regulators in HCC was analyzed. Multi-step machine learning feature selection approaches were employed to establish a model. Thereafter, we evaluated its capacity of clinical prediction and treatment guidance.

Results: First, we depicted the prognostic stratification value of CR regulators in HCC. Two CR-related phenotypes were identified, revealing a distinct clinical outcome, biological pathways and drug sensitivity. Subsequently, via four topological approaches and differentially expressed genes (DEGs) from real-world cohorts, we screened out CRY2 as the pivotal CR regulator with significant prognostic value in HCC. We performed the relevant basic assay validation for CRY2. Overexpression of CRY2 inhibited the proliferation and migration abilities of Huh7 and Hep3B cells. Moreover, three machine learning algorithms [random forest (RF), extreme gradient boosting (XGBoost) and least absolute shrinkage and selection operator (LASSO)] were implemented to construct a risk-scoring model named CR predictor, which exhibited clinical benefits and therapeutic advantages for HCC. An online nomogram based on CR predictor was developed for predicting individualized survival (https://lihc.shinyapps.io/CR_predictor/). Finally, Mendelian randomization (MR) was performed. Among model genes in CR predictor, PPARGC1A revealed a significant causal effect on HCC.

Conclusions: We proposed a CR-related risk classifier in HCC, to predict patients' overall survival (OS) and therapeutic response. Targeting CR could be a promising treatment modality against HCC.