Principal Component Analysis of Different Economic Traits in Layer Chicken

Principal component analysis is a mathematical procedure that transforms a number of possibly correlated variables into smaller number of uncorrelated variables. It can be used for simplification of data, data reduction, data classification, variable selection and many more (Wold et al., 1987). The main function of principal component analysis is to extract the important information from the data and to express this information as a set of summary indices known as principal components. Since there is a higher demand for eggs, PCA has been successfully used to describe the economic traits (productive and reproductive traits) of chickens. PCA will improve not only bird management, but also the selection of multiple economic features and the preservation of unique biodiversity. (Yunusa et al., 2013). The main benefit of implementing principal component analysis is to minimize the variance in least square sense and maximize the variance of projection coordinates. It is one of the most prevalent types of vector analysis since it decreases the dimension of the original data set and explains its variation.

Yamaki et al. (2009) used principal component analysis to study meat type chicken production traits and concluded that principal components can be used in prediction of production traits in chicken. Paiva et al., (2010) assessed the possibility of discarding production variables in laying hens (White leghorn) by principal component analysis to eliminate unnecessary and difficult to measure characteristics and found that eight of the eleven principal components showed variance lower than 0.7 (eigenvalue lower than 0.7), suggesting eight variables to discard. And only three variables viz. egg production rate between 26^th to the 58^th week of age, individual mean weight at the 34^th week of age and egg mean weight at the 58^th weeks of age were recommended for use in future experiments. Saikhom et al., (2018) used the principal component analysis to reconnoitre the interdependence in the original eight morphometric traits in Haringhata Black chickens at 22^nd week of age, out of which only five traits namely body weight, breast girth, keel length, body length, ornithological measurement, beak length, beak width and back length had the highest loading factors for first principal component (PC1) which explained the maximum variability of size and shape of this breed.

Punjab Red is a layer variety of chicken which is commonly used as backyard poultry in Punjab. Desi Cross 1 and 2 were developed using RIR and Punjab red with crossing of local backyards birds for the improvement in production performances. Egg production is considered as most important parameter in layer chicken. All three genetic groups produce brown shell eggs. Keeping this in view, the current study was designed to identify the main components and comprehend the relationship between various economic traits in layer chicken.

The present experiment was conducted at Poultry Research Farm of Directorate of Livestock Farms, Guru Angad Dev Veterinary and Animal Sciences University, Ludhiana from 2020-2021. Data were collected from 450 brown shelled egg producing layer birds from three genetic groups: Desi cross 1, Desi cross 2 and Punjab Red (150 birds per genetic group). Desi cross1 was developed by crossing of Rhode Island Red with local backyard birds, Desi cross 2 was developed by crossing of Punjab Red with local backyard birds of Punjab state and Punjab Red was previously developed and maintained at the poultry research farm. The birds having incomplete data were removed from analysis. A total of 25 traits viz. weight from 0-20th week of age (BW 0-20) and 40th week (BW 40), body weight at first egg (BWSM), age at sexual maturity (ASM), egg production at 40th week (EP 40), egg production at 52nd week of age (EP 52) were included. All the birds were produced from single hatch and hatch effect was not taken for mean comparison of among genetic groups. All of the birds included in the study received similar feeding and management practices. Electronic weighing balance was used for the measurement of different traits. 
 
Statistical analysis
 
Using Pearson correlation the phenotypic correlation was estimated between different economic traits (productive and reproductive traits). The highly correlated traits were subjected to a multivariate PCA. Rotation of principal components was done using varimax rotation for the transformation of components to approximate simple structure. At 1% level of significance the validity of data set was established by using Kaiser-Meyer-Olkin (KMO) test of sampling adequacy and Bartlett’s test of sphericity. principal component analysis as described by Everitt et al., (2001), is a method for transforming variables in a multivariate data set, x1, x2…xn into new uncorrelated variables y1, y2,……….yn which account for decreasing proportion of the total variance in the original variables specified as
y1= a11x1+a12x2+ ……+ a1nxn.
y2= a21x1+a22x2+……+ a2nxn.
yn= an1x1+an2x2+..….+ annxn.

The principal components y1, y2…….yn account for decreasing proportions of the total variance in the original variables x1,x2,……xn. Variance maximizing orthogonal rotation was used in thelinear transformation of the factor pattern matrix in order to make the interpretation of the extracted principal components easier. The principal components analyses were performed using the factor program of SPSS 24 (2016) statistical package.

Economic traits
The descriptive statistics for all traits for each genetic group and pooled are shown in Table 1. Desi cross 1 performed substantially better (P<0.05) than Desi cross2 and Punjab Red for egg production traits, especially egg production at 40^th and 52^nd week of age. For age at sexual maturity Desi cross1 (156.87±1.14 days) performed better than Desi cross2 (164.76±1.73) and Punjab Red (165.37±1.51 days) as lower age at sexual maturity is desirable trait in layer chicken. Desi cross 1 may be considered for rural poultry farming in Punjab and nearby states of the country. Kiani-Manesh et al., (2002) also suggested that age at sexual maturity, number of eggs, egg weight and body weight at eight weeks of age are the most important traits for improving the economic efficiency of rural chickens.


 
Phenotypic correlations
 
The magnitude of correlation coefficient ranges from -0.25 to 0.87. The values of all the correlation were found to be positive except between BW1-20 week and ASM is negative but it is desirable because sooner the bird attains sexual maturity, it becomes more productive. There is positive correlation between BW40 and ASM but negative correlation was observed between BW40 with EP40 and EP52 because less body weight leads to low egg production. Among all correlation, five (EP40-BW1, EP40-BW2, EP40-BW3, EP40-FEW and EP40-BW14) were highly significant (P<0.01). Similar trend was depicted by Nigussie et al., (2011) for the associations between the phenotypic performance of body weight growth and total egg production at 44 weeks of age in Horro chicken. The phenotypic correlation of age at sexual maturity was positive with FEW and negatively correlated with 40-week egg production (-0.479±0.03), 52-week egg production (-0.431±0.03), 40-week egg weight (-0.058±0.04), 52-week egg weight (-0.012±0.04) and body weight at sexual maturity (-0.013±0.04). However, contrary results were observed by Liu et al., (2019) for ASM with 40-week egg production and 52-week egg production.
 
Principal component analysis
 
Principal component analysis were applied to 25 economic traits of three different genetic groups (Desi Cross 1, Desi Cross 2 and Punjab Red). The measurement of sampling adequacy was found to be 0.95 which was estimated by Kaiser-Meyer-Olkin method. This value represents that whether each factor has enough data to give reliable results for PCA. The value should more than 0.6 and desirable value should be more than 0.8 (Tabachnick and Fitell, 2013). To maximize the sum of loading squares varimax rotation method (Fernandez, 2002) was applied. The overall significance of correlation matrix was tested by Bartlett’s test of sphericity and chi-square value was highly significant (P<0.01) and estimated as 20910.58. The sum of square loadings was extracted by PCA, Eigenvalues (Fig 1) and variation explained by each component is given in Table 2. Out of total 25 components three components were extracted using Kaiser rule criterion (Johnson and Wichern, 1982) for determination of number of significant components. A total variance of 75.524% was explained by three principal components (PC1, PC2 and PC3) which were having Eigen value more than 1. The variability of individual traits was as per explained by PCA (Mavule et al., 2013). The component plot of the three components in rotated space is shown in Fig 2.

​



​

In current study, the first principal component (PC1) explained for 38.89% of the total variance. First principal component (PC1) described the body weight 10th week to 20th week (BW 10-BW 20) and BWSM in crossbred layer chicken. It was represented by a very high component loading for BW18, BW19 and BW20. Findings of Yakubu et al., (2009) revealed total variation of 85% was explained by body weight. Egena et al., (2014) found similar results with PC1 explaining 38.3% total variation. Saikhom et al., (2018) shows similar result with body weight explaining 60.2% total variation. Yakubu and Ari (2018) concluded the result that explained body weight has maximum share of total variance (92%). Vilakazi et al., (2020) revealed principal component 1 has large share on body weight.

The second principal component (PC2) explained for 27.072% of total variation with high loads on 1 week body weight to 9-week body weight (BW1-BW9). Similar results were reported by Yakubu et al., (2009), Yakubu and Ari (2018) and Vilakazi et al. (2020) in which PC2 has large share on body weight. The third principal component (PC3) explained 9.560% of total variance and had high loads on age at sexual maturity, 40-week egg production and 52-week egg production. Yamaki et al., (2009) and Savegnago et al., (2011) also shown similar results in which age at sexual maturity had its load on principal component.

Table 3 represents the coefficient of principal component analysis of rotated component matrix. The different weights were assigned by PC1 and PC2 to all the biometric traits which were having positive sign except age at sexual maturity (ASM). The PC3 gave different weights with negative sign to BW1 week, BW40 week, ASM and BWSM. All the other traits had positive sign. The communalities ranged between 0.040 (BW40) to 0.89 (EP40). The traits like BW40, BW1, BW2, BW3 and ASM had lower communality indicating that these traits are less effective in explaining the economic traits in layer chicken. From this study it was concluded that implementation of PCA can be done in breeding programme for selection of crossbred layer birds and performance trait can be predicted and evaluated using principal components.

​
 

PC1 had the largest share of overall variance and had high loading on BW18, BW19, and BW20 in distinct genetic groups of crossbred layers chicken. Similarly, PC2 was seemed to have high loads on BW1 to BW9 week. The subsequent component PC3 was found to have high loads on ASM, EP 40 week and EP 52 week. The three main components discovered in the study could be used as a factor score to predict various economic traits. These principal components moreover provide a way in dimension reduction of economic traits which can be used in breeding programme of layer chicken.

None