Flow resistance in open channels with dune bedform is a substantial issue due to the influence of dunes on the hydraulic roughness, which can affect the performance of hydraulic constructions. There are a number of nonlinear approaches that have been developed to predict the roughness coefficient in alluvial channels, such as developed equations based on the mean velocity or shear stresses. However, due to the multitude of factors influencing roughness, establishing an accurate determination of the roughness coefficient is difficult. This study applies gene expression programing (GEP) and nonlinear approaches to predict the Manning's coefficient in dune bedform channels. Four different experimental data series were used for modeling. In order to develop the models, three scenarios with different input combinations were considered: scenario 1 considers only flow characteristics, scenario 2 considers flow and bedform characteristics, and scenario 3 considers flow and sediment characteristics. The results proved that GEP is capable of predicting the Manning's coefficient. It was found that for estimation of the roughness coefficient in dune bedform channels, scenario 3 performed more successfully than others. Sensitivity analysis showed that the Reynolds number plays a key role in the modeling process. Comparisons between GEP models and existing equations indicated that GEP models yield better results.