Current Bioinformatics - Volume 20, Issue 4, 2025

Cancer has emerged as the “leading killer” of human health. Survival prediction is a crucial branch of cancer prognosis. It aims to estimate patients' survival risk based on their disease conditions. Accurate and efficient survival prediction is vital in cancer patients' treatment and clinical management, preventing unnecessary suffering and conserving precious medical resources. Deep learning has been extensively applied in cancer diagnosis, prognosis, and treatment management. The decreasing cost of next-generation sequencing, continuous development of related databases, and in-depth research on multimodal deep learning have provided opportunities for establishing more functionally rich and accurate survival prediction models.

The current area of cancer survival prediction still lacks a review of multimodal deep learning methods.

We conducted a statistical analysis of the relevant research on multimodal deep learning for cancer survival prediction. We first filtered keywords from 6 known relevant papers. Then, we searched PubMed and Google Scholar for relevant publications from 2018 to 2022 using “Multimodal”, “Deep Learning” and “Cancer Survival Prediction” as keywords. Then, we further searched the related publications through the backward and forward citation search. Subsequently, we conducted a detailed analysis and review of these studies based on their datasets and methods.

We present a comprehensive systematic review of the multimodal deep learning research on cancer survival prediction from 2018 to 2022.

Multimodal deep learning has demonstrated powerful data aggregation capabilities and excellent performance in improving cancer survival prediction greatly. It has made a significant positive impact on facilitating the advancement of automated cancer diagnosis and precision oncology.

Cancer driver genes are genes responsible for cancer genesis; thus, identifying cancer-related genes is crucial in fostering cancer treatment. The accuracy in identifying cancer driver genes within the vast pool of normal passenger genes directly influences the efficacy of treatment approaches.

This research aimed to effectively identify cancer driver genes using the List-based Simulated Annealing (LBSA) optimization technique.

The proposed model (LBSA-DRIVER) harnesses a list-based simulated annealing algorithm within a bipartite network to pinpoint cancer driver genes. The process begins with creating a bipartite graph that integrates gene mutations and expression data from carefully chosen datasets. The LBSA algorithm is then applied to the generated graph to identify and rank the genes, drawing insights from a biological interaction network.

Following the algorithm's development, rigorous experimental analyses have been conducted using four benchmark datasets from The Cancer Genome Atlas (TCGA) database. The datasets used were the Breast Cancer dataset (BRCA), Prostate Adenocarcinoma dataset (PRAD), Ovarian cancer dataset (OV), and Glioblastoma Multiforme dataset (GBM).

Our findings, including precision, recall, F-score, and accuracy metrics, provide strong evidence of the effectiveness of the proposed model in identifying driver genes.

Accurately predicting survival in hospitalized COVID-19 patients is crucial but challenging due to multiple risk factors. This study addresses the limitations of existing research by proposing a comprehensive machine-learning framework to identify key mortality risk factors and develop a robust predictive model.

This study proposes an analytical framework that leverages various machine learning techniques to predict the survival of hospitalized COVID-19 patients accurately. The framework comprehensively evaluates multiple clinical indicators and their associations with mortality risk.

Patient data, including gender, age, health condition, and smoking habits, was divided into discharged (n=507) and deceased (n=300) categories. Each patient was characterized by 92 clinical features. The framework incorporated seven feature ranking algorithms (LASSO, LightGBM, MCFS, mRMR, RF, CATBoost, and XGBoost), the IFS method, and four classification algorithms (DT, KNN, RF, and SVM).

Age, diabetes, dyspnea, chronic kidney failure, and high blood pressure were identified as the most important risk factors. The best model achieved an F1-score of 0.857 using KNN with 34 selected features.

Our findings provide a comprehensive analysis of COVID-19 mortality risk factors and develops a robust predictive model. The findings highlight the increased risk in patients with comorbidities, consistent with existing literature. The proposed framework can aid in developing personalized treatment plans and allocating healthcare resources effectively.