The use of predictive models in Neotropical basins is relatively new, and applying these models in large basins is hindered by the lack of ecological, geographical, and social-environmental knowledge. Despite these difficulties, we used data from the das Velhas River basin to apply the BEAST (Benthic Assessment of SedimenT) methodology to evaluate and classify the level of environmental degradation. Our two main objectives were to modify and implement the BEAST methodology for use in biomonitoring programs of Brazilian basins, and to test the hypothesis that a gradient of environmental degradation determines a gradient in the structure and composition of benthic macroinvertebrate assemblages. We evaluated 37 sites: 8 in the main river, 15 in the main tributaries with different impact levels, and 14 in tributaries with minimally disturbed conditions (MDC). The BEAST model allowed us to classify 16 test sites: two as natural, four as altered, three as highly altered, and seven as degraded. Our results indicated degradation of the das Velhas River basin near its urban areas. The BEAST model indicated that the pollution gradient found among the sites generated a gradient of the macroinvertebrate assemblages, corroborating the hypothesis.