One experiment was run with broiler chickens, to obtain prediction equations for metabolizable energy (ME) based on feedstuffs chemical analyses, and determined ME of wheat grain and its by-products, using four different methodologies. Seven wheat grain by-products were used in five treatments: wheat grain, wheat germ, white wheat flour, dark wheat flour, wheat bran for human use, wheat bran for animal use and rough wheat bran. Based on chemical analyses of crude fiber (CF), ether extract (EE), crude protein (CP), ash (AS) and starch (ST) of the feeds and the determined values of apparent energy (MEA), true energy (MEV), apparent corrected energy (MEAn) and true energy corrected by nitrogen balance (MEVn) in five treatments, prediction equations were obtained using the stepwise procedure. CF showed the best relationship with metabolizable energy values, however, this variable alone was not enough for a good estimate of the energy values (R² below 0.80). When EE and CP were included in the equations, R² increased to 0.90 or higher in most estimates. When the equations were calculated with all treatments, the equation for MEA were less precise and R² decreased. When ME data of the traditional or force-feeding methods were used separately, the precision of the equations increases (R² higher than 0.85). For MEV and MEVn values, the best multiple linear equations included CF, EE and CP (R²>0.90), independently of using all experimental data or separating by methodology. The estimates of MEVn values showed high precision and the linear coefficients (a) of the equations were similar for all treatments or methodologies. Therefore, it explains the small influence of the different methodologies on this parameter. NDF was not a better predictor of ME than CF.
bird; wheat; metabolizable energy; prediction equation
