This study explores the complex interactions between sediment texture, dissolved organic carbon (DOC) levels, and water chemistry in influencing the solubility and mobility of toxic metals (Cd, Ni, Zn, Cu, Cr, Pb) in river sediments. A multi-tiered approach integrating machine learning, Structural Equation Modeling (SEM), and geochemical simulations was employed to understand metal behavior in the Meghna River, Bangladesh. Redundancy Analysis (RDA) revealed that sediment texture and DOC fractions are the primary drivers of metal mobility, with clay content contributing the most to variation in metal concentrations (Variance Inflation Factor (VIF) values for clay = 3.50). The study employed Random Forest (RF) and XGBoost models to predict metal concentrations, achieving exceptional predictive accuracy with Area Under the Curve (AUC) values of 1.000 for Ni, Zn, Cr, and Pb, and 0.964 for Cd. Regression models demonstrated strong performance with R2 values of 0.963 for Pb, 0.938 for Ni, and 0.928 for Zn, highlighting the robustness of DOC and sediment texture in explaining metal variability. SEM analysis indicated that pH mediates the DOC–metal relationship, with standardized path coefficients for DOC retention and metal mobility being −0.475 and 0.96 for Zn, respectively. The GIS-based Metal Mobility Index (MMI) and Soil Mobility Index (SMI) predicted high-risk zones for metal mobility, with an AUC of 0.91, effectively distinguishing between high and low mobility regions. These findings provide critical insights into metal transport dynamics and offer valuable tools for river sediment management and metal contamination risk assessment.