A Predictive Model for Anemia and Coronary Heart Disease Based on Bidirectional Two-Sample Mendelian Randomization and Machine Learning.

Abstract

Introduction: Anemia has been linked to an increased risk of coronary heart disease (CHD), yet the underlying causal relationship remains unclear. This study aimed to investigate the bidirectional associations between anemia and CHD using a multi-method approach.

Methods: Data were obtained from the European FinnGen biobank and the Gene Expression Omnibus (GEO) database. Mendelian Randomization (MR) analysis was performed with instrumental variables (IVs). The study assessed causal robustness using MR methods and sensitivity analysis, followed by differential expression analysis and weighted gene co-expression network analysis (WGCNA) to screen for core genes. Further, machine learning algorithms, such as least absolute shrinkage and selection operator (LASSO), random forest (RF), and support vector machine (SVM) algorithms, were applied to screen for key diagnostic genes. Additionally, the CIBERSORT algorithm was used to analyze immune cell infiltration, and validation was conducted using an in vitro oxidized low-density lipoprotein (ox-LDL)-induced endothelial cell model and western blot experiments.

Results: MR analysis revealed a positive causal link among vitamin B12 deficiency anemia, hemolytic anemia, and coronary heart disease, while cardiovascular events appeared to have a negative association with hemolytic anemia. Integrated bioinformatics analysis identified six core genes involved in immune response, inflammation, and lipid metabolism. To improve the accuracy of key gene screening and avoid bias from a single method, this study combined multiple machine learning algorithms for comprehensive analysis, ultimately identifying IFIH1 and APBB2 as potentially valuable diagnostic biomarkers, and revealing affected macrophages, mast cells, and T cells infiltration. In vitro experiments confirmed altered expression of IFIH1 and APBB2 upon ox-LDL treatment, supporting their role in CHD pathogenesis.

Conclusion: This study, through the integration of MR, transcriptomics, and machine learning methods, has for the first time revealed the causal role of vitamin B12 deficiency anemia and hemolytic anemia in the occurrence of CHD, and identified IFIH1 and APBB2 as potential biomarkers. This research study has provided a new theoretical basis and research direction for understanding the molecular link between anemia and CHD and for improving clinical early warning systems.