### INTRODUCTION

_{n}) of DDGS (Batal and Dale, 2006; Fastinger et al., 2006). Color score was found to be a fast method to determine corn DDGS quality (Ergul et al., 2003). A previous study has found that lighter colored DDGS had a greater concentration of bioavailable lysine (Pahm et al., 2009). In addition, some regression equations have been established to predict energy utilization based on chemical composition for poultry. Several equations were established to estimate the TME

_{n}based on proximate composition, however, the accuracy of the prediction equation was low (Batal and Dale, 2006). And the color scores included in the model also improved the equation for wheat DDGS (Cozannet et al., 2010).

_{n}predicting equation for cecectomized roosters based on chemical composition and color scores.

### MATERIAL AND METHODS

### Samples, animals and experimental design

_{n}and true digestible amino acid (TDAA) of DDGS. From periods 1 to period 3, seven DDGS samples were randomly assigned to each group with 5 replicates (1 bird per replicate). The other four samples were studied in period 4 with 5 replicates (1 bird per replicate) per group. In each period, birds were given 50 ml of 60% glucose solution after a 24 h fast. After a further 24 h, each bird was accurately tube fed 30 g of DDGS. All excreta were collected for the subsequent 48 h. Excreta samples were weighted, 10% HCl (1 ml/10 g excreta) added and stored at −20°C. After sample collection, all birds were fed a conventional corn-soybean meal diet for approximately 10 to 14 d to recovery body condition and randomly assigned to each group based on similar body weight before next period. Another 2 periods were used to estimate individual bird endogenous AA and energy losses. All roosters were fasted for 24 h and then perfused with a glucose solution. The excreta were collected for 48 h as described.

### Samples analyses

### Calculations

_{n}was calculated according to the procedure outlined by Farhat et al. (1998). where GI is the gross energy intake (kcal), GO is gross energy output (kcal), FI is feed intake of the DDGS (30 g), FGL is the fasting gross energy loss (kcal) from the feed deprived birds, ANR is apparent nitrogen retention (g), and FNL is fasting N loss (g). The gross energy excreted was corrected to zero-N balance using a factor of 8.22 kcal/g. Non-amino acids CP (NAACP) was in DDGS samples were calculated by CP minus total AA.

### Statistical analysis

_{n}to GE) were more than 2 SD above or below the mean. Two samples were removed from the data set. The regression analysis was performed by the new database. Prediction equations of TME

_{n}with individual proximate compositions and color scores were established by regression analysis with stepwise method (using PROC Reg of SAS). The best fit equations were defined by Root MSE, Mallows statistic C(p) and BIC. If the intercept was not significant (p>0.15) in the equation, it was removed from the model and the R

^{2}was calculated using the NOINT option of SAS.

### RESULTS

### Nutrient composition and color score

_{n}ranged from 1,172 to 2,782 kcal/kg and 17.1% of CV. Among all proximate compositions, ADF had the highest variation (CV = 23.5%), and DM had the lowest variation (CV = 2.2%). CV for the other nutrients ranged from 4.9% to 15.9%. EE content ranged from 9.80 to 19.40% and averaged 15.22%. Values of NAACP between 0.14 and 11.82% and 55.0% of CV. CV for L*, a* and b* were 12.5%, 9.1% and 13.3%, respectively.

### Correlation between variables

_{n}was positively correlated with GE and EE (p<0.05) and negatively correlated with CF (p<0.01). GE was (p<0.05) positively correlated with EE and color scores. There were no significant (p>0.10) correlations of color scores with ash, CP, CF, NDF, ADF and NFE. Ether extract was positively correlated with b* value (p<0.05) and negatively (p<0.05) correlated with NFE.

### Prediction equations

_{n}of 23 DDGS samples (Table 5). GE was the first independent variable included in the model in equation 1 (R

^{2}= 0.65; BIC = 249.3; RMSE = 199.3). The model was improved with the addition a* in equation 2 with a R

^{2}of 0.69. And the best fit equation was: TME

_{n}, kcal/kg = −2,995.6+0.88×GE+49.63×a* (R

^{2}= 0.69; BIC = 248.8; RMSE = 190.8) at DM bases. After removing GE from the model, the best single factor to predict TME

_{n}was b* value (equation 3; RMSE = 259.5), followed by EE and a*. When a* was included in the model, b* was discarded (p = 0.69). In this condition, the best fit model was: TME

_{n}, kcal/kg = 57.88×EE+87.62×a* (BIC = 254.3, RMSE = 223.5) at DM bases.

### DISCUSSION

_{n}precisely (R

^{2}= 0.34; RMSE = 361.9, data not shown), so the outliers were removed from the data set for further analysis. Because of the high correlation between GE and TME

_{n}and the ratio of TME

_{n}to GE being relatively stable, the outliers were removed when the ratio (TME

_{n}to GE) was below or above 2 SD of the average value. Unlike the previous report, no proximate composition parameters were included in the model when gross energy and a

^{*}were in the equation. Rochell et al. (2011) reported that nutrients including NDF, GE and CP can be used to generate AME

_{n}prediction equations for corn co-products (R

^{2}= 0.87), but they did not observe a correlation between AME

_{n}and GE (r = 0.214, p = 0.44). Gross energy was strongly correlated with TME

_{n}in the present study.

_{n}was EE and followed by CF, CP and ash, but these equations may only be used as a general guide (R

^{2}= 0.45). In contrast to the results of the current study, Cozannet et al. (2011) predicted AME

_{n}of wheat DDGS with L* (R

^{2}= 0.77) instead of a*, while this study did not determined a* or b* values.

_{n}showed considerable variation among samples (CV = 17.1%, n = 25) and the range spaned 1,610 kcal/kg. Small variation in TME

_{n}values of DDGS, however, was reported by Parsons et al. (2006), who determined 20 DDGS on samples from feed mills in Minnesota with a CV of 3.6%; another study (Batal and Dale, 2006) determined 17 DDGS samples from 6 different plants in the Midwestern United States with the TME

_{n}ranging from 2,490 to 3,190 kcal/kg (86% DM basis) and a CV of 6.4%. Although DDGS samples were high in fat, lower TME

_{n}(Average = 2,144 kcal/kg, 46% of GE) was observed in the present study compared to the values in the United States in which TME

_{n}takes nearly 60% of GE (Fastinger et al., 2006; Kim et al., 2010). One possibility of low TME

_{n}is that DDGS in the present study were high in NDF, which cannot be digested by poultry (except turkey), and a high fiber diet enhanced intestinal movement that reduced feed retention time and feed digestibility. It is well known that most starch and sugar in DDGS were fermented to ethanol, and less than 7.0% starch is in the DDGS (Belyea et al., 2004; Stein et al., 2006). NFE is considered as a measure of the digestible carbohydrates (Leeson and Summers, 2009), thus another possible contributor to a low TME

_{n}could be the low concentration of NFE in DDGS samples.

_{n}. In addition, gross energy is the best single indicator of TME

_{n}. Finally, in this study, color score cannot predict TDAA or TD lysine of DDGS.