Animal numbers per breed, sex, examination year, weight after examination were displayed in Table 1.

Data used for this study (n = 2,495) were collected from three pig breeding farms in Korea. Four breeds, Berkshire (n = 1,345), Duroc (n = 459), Landrace (n = 245) and Large White (n = 446), of farm performance tested pigs were measured for IMF percentage using live animals and their pork quality trait records were collected.

At an average weight of 114 kg, pigs were scanned with an Aloka 500V SSD ultrasound machine (Corometrics Medical Systems, Inc., Wallingford, CT, USA) to measure their IMF percentage. A minimum of four longitudinal images (Figure 1) were collected at 7 cm off the midline across the 10th to the 13th ribs. The images were used to predict UIMF in accordance with the method proposed by Newcom et al. (2002).

Following a pH_24h (24 hour chill after harvest, the ultimate pH) was measured on the 10-rib face of the longissimus muscle using a pH star probe (Horiba 6252-10D, Irvine, CA, USA). At 24 h postmortem, a 3.2 mm slice of the loin beginning from the 10th- to the 11th-rib interface was analyzed for CIMF using Bligh and Dyer’s method (1959). The frozen samples were thawed overnight at 4°C and trimmed, with 5 mm cut off the top to expose a fresh surface. The objective meat color (Commission Internationale de l’Eclairage [CIE]; lightness [L*], redness [a*], yellowness [b*]) was measured after a 30 min blooming period at 4°C (CR-300, Minolta Camera Co, Osaka, Japan). Cooking loss (CL) and shear force (SF) measurements were obtained from approximately 350 g (5×6×7 cm) of meat blocks, trimmed and cooked in a 70°C water bath for 60 min until the sample block reached 70°C. The cooked samples were cooled in 18°C running water for 30 min. The SF was determined from an average of 6 cores taken from the cooked sample block using an Instron (Series IX, Instron Corp, Norwood, MA, USA). The measurement condition was 400 mm/min head speed with 40 kg load cells. Weight loss before and after cooking was expressed as a percent weight of the initial cooking block and ascribed to CL.

To estimate the effects of breed, sex, final weight, parity, birth year and birth month on productive traits, the collected data were statistically analyzed by the least squares method using the following linear model (Harvey, 1979).

Where, Y_ijklm is the observation on the i^th breed, j^th sex, k^th year of test and l^th final weight; μ is the population mean; B_i is the effect of i^th breed (i = 1, 2, 3, 4); S_j is the effect of j^th sex (j = 1, 2); Y_k is the effect of k^th year of test (k = 1, 2, 3); W_l is the effect of l^th final weight (l = 1, 2, 3, 4, 5); e_ijklm is the residual error.

The restriction imposed to solve the normal equation was to assume the last effect within each classification to be zero.

All parameters were analyzed using the generalised linear model (GLM) procedure in SAS ver 6.12 (SAS Institute, 1996). The PROC GLM procedure was used for analysis of variance and the treatment group was set as the class variable. The least squares means (LSM) for each treatment group was computed for group comparison (H0: LSMi = LSMj) using the LSMEANS option in PROC GLM. The PROC CORR procedure was used to compute the phenotypic correlations between meat quality and UIMF percentage. Statistical significance was taken to be 5% (p< 0.05).

The following linear model was used to estimate the variance and covariance components of the parameters studied:

where, Y is a vector of observations ; β is a vector of fixed effect (sex, year of test, final weight); u is a vector of random additive genetic effects; Z and X are known incidence matrices relating observations to the respective fixed and random effects; and e is a vector of random residual effects.

A derivative-free restriced maximum likelihood algorithm, as applied in the WOMBET programs (Meyer, 1991), was used to estimate components of variance. Solutions for fixed effects and estimated breeding values were obtained from these when convergence was attained. Iterations were stopped when variance of function values in the simplex were less than 1×10⁻¹⁰. Each analysis was then restarted using the calculated estimates as new priors until changes in the function values were observed.

Least squares means of the meat quality, UIMF percentage for breed and their respective standard errors are presented in Table 2 (p<0.01).

Objective meat quality and palatability are greatly influenced by the pH of meat due to the effect of pH on water-holding capacity, protein solubility and surface reflectance (Joo et al., 1999; Hwang et al., 2003). Significant differences in ultimate pH at 24 h postmortem were observed in the different breeds used in this study. The ultimate pH of the Berkshire and Duroc breeds at 24 h postmortem were higher than the Large White and Landrace breeds (p<0.05). This result supports the observation by Newcom et al. (2004) and Stalder et al. (2003) that Berkshire and Duroc breeds had higher pigment content than Landrace and Yorkshire breeds. Berger et al. (1994) reported that Hampshire breeds had lower (p<0.05) pH than any other breed.

Several studies have demonstrated that the level of IMF percentage greatly influences eating quality (Wood et al., 1996). Significant breed differences for CIMF and UIMF were observed in this study, with the Duroc breed having higher IMF than the Berkshire, Landrace, and Large White breeds (p<0.05). These results are in agreement partly with the report by Newcom et al. (2004) that Landrace and Yorkshire breeds had higher pigment content than Berkshire and Duroc breeds. The Duroc and Berkshire breeds had the highest (p<0.05) levels of IMF percentage, which agrees with the findings of Berger et al. (1994).

Significant breed differences for CIE L* were also observed in this study. The Landrace and Large White breeds had higher CIE L* than the Berkshire and Duroc breeds (p<0.05). This result is also in agreement with the study by Newcom et al. (2004). Additionally, the Duroc, Landrace, and Large White breeds had higher CIE a* and b* than the Berkshire breed (p<0.05) while the Landrace and Large White breeds had higher CL and SF than the Berkshire and Duroc breeds (p<0.05).

Genetic parameter estimates for meat quality traits and UIMF percentage are displayed in Table 3. The heritability estimate for ultimate pH was determined to be 0.33, which is higher than the value reported by Larzul et al. (1997) and Suzuki et al. (2005). The heritability estimates for CIMF and UIMF were calculated to be 0.50 and 0.48, respectively. Both values were higher than the values reported by Newcom et al. (2005). The calculated CIE L*, a*, and b* values of 0.51, 0.50, and 0.48, respectively, were lower than the values reported by Newcom et al. (2003). A previous review by Sellier (1998) determined that the average heritability for CIE L* was 0.28 using data from 29 published estimates with a range of 0.15 to 0.57, which was lower than the estimate found in this study. The CL and SF were determined to have moderate heritability (0.46 and 0.56). Suzuki et al. (2005) also reported the heritability for drip loss and CL as 0.14±0.01 and 0.09±0.02, which was lower than the estimate found in this study.

Genetic and phenotypic correlations between ultimate pH and CIMF were −0.15 and −0.02, respectively. A higher negative genetic correlation of −0.51 was reported by Suzuki et al. (2005). Genetic and phenotypic correlations between ultimate pH and UIMF were −0.14 and −0.01, respectively. Genetic and phenotypic correlations between ultimate pH and meat color were negative, in agreement with findings by Suzuki et al. (2005). This relationship is also supported by other reports (DeVol et al., 1988; Hovenier et al., 1992). Conversely, pork with a low ultimate pH can be lighter in color (Monin and Sellier, 1985).

Estimates of genetic and phenotypic correlations for CIMF and UIMF were 0.75 and 0.76, which were strong correlations. This correlation suggests that CIMF can be replaced with UIMF and that ultrasound machines can be used to determine IMF in live swine.

In this study, the range of genetic correlations between IMF (CIMF and UIMF) and CIE L*, CIE a*, CIE b* were 0.24 to 0.27, 0.30 to 0.37, and 0.41 to 0.48, respectively. These results are similar to results reported by Suzuki et al. (2005). The range of genetic correlation between IMF (CIMF and UIMF) and CL was determined to be 0.01 to 0.03. This was in agreement with Hovenier (1992), but not with Suzuki et al. (2005). Genetic and phenotypic correlations between meat color (CIE L*, CIE a*, CIE b*) and CL were 0.13 to 0.23 and 0.28 to 0.29, respectively. The genetic and phenotypic correlations between CL and SF were estimated to be 0.24 and 0.33, respectively.