Multiple linear regressions equations were developed to predict the chemical composition of the empty body of Nellore cattle. Thirty-tree young bulls (Control Nellore, n = 11; Selected Nellore, n = 22) of 339 ± 35 kg initial weight and 448 ± 26 d of age were used. The characteristics used to develop the prediction models were: empty body weight (EBW), age (days), hip height (HH), and ultrasound measurements [backfat thickness (uBF), rump fat (RF) and longissimus muscle area (uLMA)]. Ultrasound measurements were collected at the end of the trial period (d 110). Chemical composition (protein, ether extract, water and ash) was performed on samples obtained after grinding and homogenization of all body tissues. No interaction (P > 0.05) was observed between chemical components and genetic groups. Mallows Cp values were close to the ideal value of number of independent variables in the prediction equations plus one. Precise predictions (R² > 0.72) were obtained for protein, ether extract, and water, whereas poor prediction for ash (R² = 0.67) was revealed. Protein, ether extract, and water percentage can be calculated using the following equations: % Protein = 52.011 − 1.415 × RF + 0.179 × EBW − 32.122 × HH (R² = 0.91; SE = 2.24); % Ether Extract = 11.300 + 2.096 × RF + 0.157 × EBW − 0.061 × age (R² = 0.72; SE = 4.47); % Water = −34.156 − 2.327 × RF + 0.550 × EBW + 0.103 × age (R² = 0.96; SE = 5.81). In conclusion, multiple linear regressions were detected between non-invasive measurements and chemical composition of the empty body. The equations developed can be used to estimate chemical body composition characteristics as protein, ether extract and water body percentages of Nellore cattle. Supported by FAPESP Process#2005/60042–2.