A Machine Learning Framework for Cancer Prognostics: Integrating Temporal and Immune Gene Dynamics via ARIMA-CNN

Area of Science:

• Oncology
• Immunology
• Bioinformatics

Background:

• Hepatocellular carcinoma (HCC) presents a significant global health burden with high mortality.
• The tumor immune microenvironment (TME) critically influences HCC progression and patient outcomes.
• Traditional analyses of TME often focus on single genes, neglecting complex immune gene interactions.

Purpose of the Study:

• To investigate the prognostic value of chemokine (C-C motif) ligand 5 (CCL5) and associated immune genes in HCC.
• To develop and validate an innovative Autoregressive Integrated Moving Average (ARIMA) and Convolutional Neural Network (CNN) framework for survival analysis.
• To compare the efficacy of the ARIMA-CNN model against traditional prognostic methods.

Main Methods:

• Time series analysis of CCL5 expression in 230 HCC patients using ARIMA.
• Integration of ARIMA residuals and immune gene expression data as input for a CNN model.
• Prognostic evaluation using Cox proportional hazards models, Kaplan-Meier curves, and hierarchical clustering.

Main Results:

• The ARIMA-CNN framework demonstrated superior prognostic capability over median-based stratification.
• Features from CD8+ T cells and effector T cells showed significant association with improved survival (HR: 0.7324, p=0.0008).
• A cluster including B cells, Th2 cells, T cells, and NK cells indicated a protective effect (HR: 0.8714, p=0.1093).

Conclusions:

• The study presents the first ARIMA-CNN framework for gene expression and survival analysis, integrating temporal dynamics and machine learning.
• This integrative approach provides deeper insights into the HCC TME.
• The findings highlight potential for advancing precision immunotherapy and identifying novel biomarkers for HCC management.