Identification of Potential Biomarkers in Acute Myeloid Leukemia Using Integrated Bioinformatics and Two-Step ML-Based Feature Selection Approaches

Acute Myeloid Leukemia (AML) is a highly heterogeneous hematological malignancy with poor prognosis, necessitating the identification of reliable biomarkers for early diagnosis and targeted therapy. In this study, we employed an integrative approach combining machine learning and bioinformatics to identify potential biomarkers in AML. Differentially expressed genes (DEGs) were initially identified from a microarray dataset. Feature selection was performed using Minimum Redundancy Maximum Relevance (mRMR) and Support Vector MachineRecursive Feature Elimination (SVM-RFE) to refine the most significant genes. A protein-protein interaction network (PPINet) was then constructed using the STRING database and analyzed in Cytoscape. Utilinzing the Degree method, the top ten hub genes (CD3D, TYROBP, PIK3R1, ITGAX, CD247, CD79B, CCL5, LCK, CD8A, and CD2) were identified. Further gene regulatory network analysis, including transcription factor (TF)-gene interactions and gene-miRNA interactions, provided insights into the regulatory mechanisms of these hub genes. Additionally, druggene interaction analysis revealed potential therapeutic targets among the identified biomarkers. These findings suggest that the identified hub genes may serve as promising candidates for AML diagnosis and treatment.