Optimizing Prognostic Predictions in Liver Cancer with Machine Learning and Survival Analysis

Area of Science:

• Genomics
• Oncology
• Bioinformatics

Background:

• Liver hepatocellular carcinoma (LIHC) is a significant global health concern with poor prognosis and limited therapeutic options.
• Identifying reliable genetic markers is crucial for understanding LIHC progression and developing targeted treatments.

Purpose of the Study:

• To identify pivotal genetic markers associated with LIHC progression using RNA sequencing data.
• To evaluate the performance of various feature selection and survival analysis methods for predicting LIHC outcomes.

Main Methods:

• Utilized RNA sequencing data from The Cancer Genome Atlas (TCGA) for LIHC.
• Employed feature selection techniques: Sure Independence Screening (SIS) with Least Absolute Shrinkage and Selection Operator (Lasso), Smoothly Clipped Absolute Deviation (SCAD), Information Gain (IG), and Permutation Variable Importance (VIMP).
• Applied survival analysis models: Cox proportional hazards model, survival tree, and random survival forests.

Main Results:

• Identified MED8 as a critical gene marker for LIHC.
• SIS-Lasso combined with the Cox proportional hazards model showed strong predictive accuracy.
• The SIS-VIMP approach with random survival forests achieved the highest overall performance in predicting LIHC outcomes.

Conclusions:

• Advanced feature selection and survival analysis methods effectively identify genetic markers for LIHC.
• The SIS-VIMP approach with random survival forests offers superior predictive power for LIHC.
• Findings provide insights into LIHC's genetic underpinnings, aiding personalized medicine and cancer genomics research.