Published on
Cancer is a major cause of global mortality, necessitating accurate long-term survival (LTS) predictions for patients to optimize treatment strategies and improve prognosis. The advent of high-throughput technologies has facilitated the study of biological systems through multiple omics data, including genomics, transcriptomics, proteomics and other omics. Consequently, multi-omics approaches have emerged as promising tools for enhancing cancer biology understanding and enabling more precise LTS predictions. This study aims to assess the performance of various benchmark models (including our own in-house developed G2PDeep multiCNN model and other methods) in multi-omics approaches for LTS prediction across 23 distinct cancer studies. We compared single, dual, and triple omics methodologies to determine the most effective strategies for predicting LTS in a diverse array of cancer types. Additionally, we evaluated the generalizability and robustness of these models to identify potential limitations and areas for improvement. Our comprehensive assessment of benchmark models in multi-omics approaches will contribute to the development of more accurate and dependable predictive tools for cancer patients. Ultimately, these findings will guide clinical decision-making, facilitating the provision of more personalized and effective treatment options, and subsequently improving the overall prognosis for cancer patients.