Principal component regression of carcass traits in meat line funaab alpha chicken genotype

This study aimed to assess the relationship among carcass traits of meat line FUNAAB Alpha chicken genotype, to identify the components that dened bled weight in them using multivariate principal component regression. A total of 14 different carcass traits from sixty-eight birds were recorded and subjected to one-way analysis of variance to vet for sex effect. Phenotypic relationships among the carcass traits were also established to pave way for the principal component analysis. The results reveal signicant effects between the traits measured. The male signicantly (P<0.05) had greater mean values for the traits measured. Correlations among the considered carcass traits were found to be positive and signicant ranging from r = 0.406 (LrWt) - 0.981 (EdWt) for the female chicken; r = 0.330 (Head Wt) - 0.978 (BdWt) for the male chicken. The extracted components PC1 to PC7 contributed 95.66% with PC1 accounting for 68.68% of the variability in the original parameters. Communality estimates varied from 0.466 (thigh weight) to 0.983 (liver weight). In the principal component regression models, Eviscerated weight accounted for 95% of the variation observed in bled weight. The use of PC1 as a single predictor, explained 96.4% of the variability, whilst combining PC1 and PC4 showed improvements in the variance explained (R 2 = 96.7%) with a lower Mallow's cp (5.31). Using the principal components scores from the chicken morphometric traits was more appropriate than using the original traits in bled weight prediction.


Introduction
Poultry production in developing countries is a fast growing sub-sector of agriculture that is quick in terms of shorter production cycles (Magothe et al., 2012;Mottet and Tempio, 2017;Alabi et al., 2019;Alabi et al., 2020). Products from poultry are one of the most popular options available in Nigeria in reducing protein de ciency and malnutrition in the populace diet (Obasoyo et al., 2005;Anyawu and Okoro, 2006;Ojedapo et al., 2015). The call for the improvement and development of chicken breeds that are indigenous to Nigeria led to the development of the FUNAAB Alpha chicken breed (Broiler and dual purpose). The breed was developed at the Federal University of Agriculture, Abeokuta, Nigeria, using the Naked and Frizzled feather chicken genotype (Broiler type) having over six generations through cross breeding with some exotic lines to maximise growth and their productive performance (Adebambo et al., 2018). This development has been effected for improved production without compromise to its climatic adaptation and diseases susceptibility (Saleh et al., 2017).
The growth process within the indigenous chicken is a complex one that relates to increasing body cells and volume. Growth mechanisms and processes are mostly too multifaceted to be explained with univariate analysis. This can be because of the biological linkage caused by gene loci and pleiotropic effect. The principal component analysis which is a multivariate procedure could better solve the problems associated with the univariate analysis of growth and its observed traits. This is because it reduces related variables into a lesser numbers of uncorrelated variables called principal components (Kor et al., 2006;Rosario et al., 2008;Amao, 2018). The principal component analysis is a tool for exploratory data and is adapted for making predictive models (Acha, 2012). It has been used to describe the inter-relationship between body measurements and size in chickens (Yakubu et al., 2009a;Udeh and Ogbu, 2011;Akporhuarho and Omoikhoje, 2017). In animal genetics and improvement, principal components simultaneously considers certain attributes that may be utilised for selection purposes. An important aspect is that every principal component explains a certain percentage of the total variance with the rst principal component explaining the highest percentage of this variance (Pinto et al., 2006). This study was aimed at analyzing the principal component regression of carcass traits of meat line FUNAAB Alpha Chicken Genotype. The outcome could aid the management and subsequent selection towards genetic improvement of the indigenous stock.

Experimental site
The study was conducted at PEARL (Program for Emerging Agricultural Research Leaders) Poultry Breeding unit of the Directorate, University Farms (DUFARMS), Federal University of Agriculture, Abeokuta (FUNAAB), Ogun State, Nigeria. The area lies in the South-western part of Nigeria, it is located within the Longitude 7 0 10 0 N and Latitude 3 0 2 0 E, with an average temperature of 33.7 0 C and relative humidity of 80% with rainfall of about 1037mm (AGROMET, FUNAAB, 2014).
Experimental birds 68 birds comprising of 44 males and 24 females randomly selected from the meat line of the improved indigenous chicken, FUNAAB Alpha was used in the experiment. They were generated through arti cial insemination at PEARL Farm, Federal University of Agriculture, Abeokuta, Ogun State, Nigeria.

Feeds and feeding
Feeding was ad libitum; the chicks were on a starter diet that supplied 24% crude protein and 2900Kcal/KgME metabolizable energy from 0 -4 weeks old and a nisher diet that supplied 23% crude protein and 3000Kcal/KgME metabolizable energy from 5 -8 weeks. Clean water was given ad libitum throughout the experimental period.
Data collection: The birds were fasted for 12 hours and slaughtered by severance of the carotid arteries and jugular veins, the blood was drained under gravity and scalded to facilitate plucking and evisceration. The carcass were divided into parts as described by Kleczek et al. (2006). Data were collected on the following traits using a 0.1g sensitive scale: bled weight, de-feathered weight, eviscerated weight, breast weight, thigh weight, back weight, shank weight, wing weight, neck weight, head weight, drum stick weight, liver weight, whole gizzard weight, empty gizzard weight.
The varimax rotation maximized the sum of variances of the aij2 quadratic weight, whilst the stepwise multiple regression procedure was used to obtain models for the prediction of bled weight from the carcass traits (i) and from the established principal components (ii).
Where; BdWt is Bled weight, a is regression intercept, Bi is the ith regression coe cient of the ith linear bled weight measurement, Xi is the ith principal component.
The components were extracted until certain stopping criteria were encountered or when the p components were formed. The weights used for the formulation of the principal components were the eigenvectors of the characteristic equation: Where R is the correlation matrix, The λ1 are the eigenvalues (variances of the components).
Cumulative proportion variance were employed to determine the number of principal components for extraction. The means, correlation and regression procedures of S.A.S 9.1 statistical package was used for the principal component regression (Adenaike et al., 2015). Mallow's Cp was used to select maximal accurate subsets of predictor variables in the stepwise regression (Liberman and Morris, 2008).

Results
The means, standard error and coe cients of variation for the carcass traits of meat line of FUNAAB Alpha chicken genotype at 8 weeks of age are presented in Table 1. The results reveals signi cant effects amongst the traits measured. The male signi cantly (P<0.05) had greater mean values for the traits measured; from bled weight of 1174 ± 20.49 compared to the female with 1030 ± 29.5 to liver weight 23.69 ± 0.56 and female 20.58 ± 0.68.   The scree plot is shown in Fig. 1. The scree plot is a criterion for selecting the actual number of components which would be retained for further analysis. The components having eigenvalue up to the bent of elbow in scree plot are generally considered (Dalal et al., 2020). Only the rst two components have eigenvalues greater than 1. This accounts for 78% of the total variance. However, there is a small drop between components 2, 3 and 4. Components 5 through 14 appeared at the bottom of the plot. Component 1 to 7 accounts for 95.66% of the total variance.
The eigenvalues and shares of variance along with the rotated factor loading and communalities for the carcass traits of the meat line of FUNAAB Alpha Chicken Genotype at 8 weeks old are presented in Table 3. The communalities represents the estimates of variance for each variable accounted by the components. This ranged from 0.113 -0.983 in the traits. The eigenvalues showed the number of variances out of the total variance explained by each factors. Seven principal components PC1 to PC7 was extracted. PC1 to PC7 contributed 95.66% while PC1 accounted for 68.68% of the total variance. Furthermore, the varimax rotation method of principal component analysis indicated that the carcass traits that contributed signi cantly to PC1 were bled weight, eviscerated weight, drumstick weight and wing weight, while PC2 displayed that whole gizzard weight and empty gizzard contributed to the total variance. The communalities obtained for some traits were at high range. The highest was from liver weight while the lowest was from bled weight. The original inter-dependent carcass traits and the independent principal components (orthogonal) was used to determine bled weight of the chicken. This is seen in Table IV. Eviscerated weight accounts for 95% of variation observed in bled weight while inclusion of empty gizzard weight showed the same proportion of the explained variance, but with a lower Mallow's cp (17.11) compared to (26.73) in the rst model. The model's accuracy was further improved R 2 = 96% when neck weight was added to the equation. The best prediction equation (R 2 = 96.3%) was however obtained with a combination of eviscerated weight, neck weight, empty gizzard weight and liver weight. The highest single contributor (R 2 = 95%) to the variation in bled weight of meat line FUNAAB Alpha chickens was eviscerated weight. In this present study, PC1 as a single predictor explains 96.4% of the total variability in bled weight of meat line FUNAAB Alpha chickens. However, a combination of PC1 and PC4 shows an improvement in the total variance explained (R 2 = 96.7%) with a lower Mallow's cp (5.31)

Discussion And Conclusion
Diverse factors which includes genetics, age, live weight and sex affects meat yield, composition and quality (Khan et al., 2018). The present results showed sex associated differences in the carcass traits measured, with higher values found in the male FUNAAB Alpha chicken genotype. The apparent sex-related differences could be attributed to the average sex differential hormonal effects on the growth (Yakubu et al., 2009a), and sexual dimorphism (Baeza et al., 2001) of chicken. According to Van der Heide et al. (2016), genes for important economic traits may be differently expressed in males and females. The average bled weights observed in this study are lower compared to that of Akanno et al. (2007); Udeh and Ogbu (2011), who reported 1.88kg (Arbor Acre), 1.81kg (Ross) and 1.65kg (Marshal), this however could be due to the differences in the strain of birds used.
The positive relationship amongst bled weight and the carcass traits reveals that the weight can be predicted from the carcass traits in these chickens. Similar observations were made by Ajayi et al. (2008); Akporhuarho and Omoikhoje (2017). This asserts that the morphological traits are interrelated and indicate high predictability among the variables. Similar ndings were reported by Udeh and Ogbu (2011), who opined that the correlation values revealed the pattern amongst the variables. Positive correlations of these traits suggest also that the traits are under the same gene action (pleiotropy) while selection for a trait might lead to a correlated response for the other traits (Yakubu et al., 2009a;Adenaike et al., 2015).
The PC approach reduced the fourteen variables into seven components for better description of carcass traits of FUNAAB Alpha chicken. PC1 had high loadings on bled weight, de-feathered weight, eviscerated weight and drumstick weight, this is largely associated with body size and accounts for 68% of the variations amongst the traits. Whole gizzard weight and empty gizzard weight were the major contributors to PC2, which can be associated with internal traits. Other PCs had low contributions to the total variation in the carcass traits. However, these PCs could be vital in evaluation of animals for breeding and selection purposes and these can be a criteria for improvement in the body weights of indigenous chickens (Ajayi et al., 2012;Egena et al., 2014;Vilakazi et al., 2020). In this study, using PC1 as a single predictor explains 96.4% of the total variability in the carcass traits. This is lesser than the ndings of Ajayi et al. (2012) Scree plot of eigenvalues