Type 2 Diabetes (T2D) remains a major public health challenge, demanding analytical frameworks that not only predict disease status but also reveal underlying molecular mechanisms. This study presents the Genomic Causal Framework (GCF), which integrated differential expression analysis, Causal Bayesian Networks (CBNs), and Probability Trees (PTrees) to identify key regulatory drivers and inform interpretable prediction in the diabetic-obese transcriptome. Using RNA-sequencing data from the GSE132831 dataset, we identified 146 differentially expressed genes (DEGs), stratified into upregulated and downregulated sets. Protein–protein interaction (PPI) networks constructed from these DEGs revealed functional modules related to metabolic regulation and immune suppression. Central genes included SBF1, UTY, MMP9, and CXCL8, while CBNs identified upstream regulators such as SREBF1, NPC1L1, and CXCR2. SHAP analysis further highlighted genes such as UTY, S100P, REG3A, and S100A8 as influential in prediction. This pattern aligned with transcriptomic evidence of immune downregulation in T2D. GCF achieved 93.33% accuracy and 95.23% sensitivity, with interpretability provided by causality-informed PTrees and SHAP-based attribution. By combining probabilistic reasoning with functional network analysis, GCF distinguished upstream regulators from downstream effectors, supporting its potential for biomarker discovery and mechanistic insight. The framework offered a transparent alternative to black-box models, facilitating more interpretable applications of causal inference in precision medicine for metabolic disorders.