Phenotypic Diversity of Pea Genotypes (Pisum sativum L.) Based on Multivariate Analysis

The present investigation consisting of fifty five genotypes of pea was carried out at Horticultural Research Centre, SVPUA&T, Meerut, U.P. (29°01 latitude N and 77°43 longitude E and 219.75 MSL) during Rabi season 2015-16 to evaluate the amount of genetic diversity and relatedness among the genotypes (Table 1). The experimental design was RBD with three replications. In each replication the seeds were sown in a plot of 6.0 m × 1.0 m in which row to row and plant to plant spacing was 60 cm and 10 cm respectively. All the recommended agro-practices were followed to ensure a healthy crop growth and development. Observation were recorded on five competitive plants situated under the same field condition for eleven morphological quantitative traits viz., days to 50 per cent flowering, plant height, number of first fruiting node, length of first fruiting node, number of pods per plant, length of pod, width of pod, number of seeds per pods, green pod yield per plant, seed yield per plant and shell weight per plant. The magnitude of genetic diversity among fifty five pea genotypes was determined by using D² Mahalanobis genetic distance statistics (Mahalanobis, 1936). Hierarchical clustering using Tocher’s method, as described by Rao (1952) was followed for the grouping of genotypes into distinct clusters. Partition clustering using K-means as suggested by Charrad et al., (2014) and principal component analysis were done using R software (Venables and Ripley, 2002).
 

The data obtained from the observations recorded on eleven morphological quantitative traits were subjected to the statistical scrutiny. It was evident from the analysis of variance that mean sum of squares due to 55 genotypes were highly significant for all the traits except days to 50 per cent flowering (Table 2), giving the clear picture of presence of wide spectrum of variability among the genotypes. These results were in agreement with the findings of Lal et al., (2011); Supe et al., (2013); Georgieva et al., (2016) and Kumar and Kumar (2016). Although the analysis of variance revealed sufficient variability among the genotypes, but the extent of genetic diversity present among the genotypes could not be explained, therefore, cluster analysis was performed to quantify the genetic divergence between any two genotypes or group of genotypes.
 

 
Based on the relative magnitude of their Mahalanobis D² values using Torcher’s method, all the 55 genotypes of pea under study were grouped into six clusters. The clustering patterns of pea genotypes into six clusters are presented in Table 3.  Maximum number of genotypes (14) was grouped in cluster III namely: VRP-13, VRP-375, VRP- 324, VRP-311, VRP-176, VRP-327, VRP-276, VRP-82, VRP-248, KS-228, DPP-94/8-06, Kashi Uday, Kashi Samridhi and Kashi Nandini. Whereas, cluster I and cluster V both contained ten genotypes each where, cluster I comprises of genotypes namely: VRP-3, VRP-228, VRP-320, VRP-22, VRP-122, VRP-383, VRP-402, VRP-382, VRP-145 and Kashi Mukti; and cluster V consisted of genotypes namely: VRP-69, VRP-313, VRP-73, VRP-321, VRP-16, VRP-284, VRP-223, VRP-343, VRPM-15 and Kashi Shakti. Furthermore, cluster II comprises of nine genotypes namely: VRP-26, VRP-273, VRP-107, VRP-156, VRP-174, VRP-49, VRP-131, EC-8724 and MO-23; followed by eight genotypes that were arranged in cluster VI namely: VRP-194, VRP-115, VRP-355, VRP-65, VRP-64, VP-233, EC-97280 and EC-8372. Cluster IV consisted with minimum of four  genotypes namely: VRP-222, VRP-95, EC-71944 and MO-19. No parallelism was shown by the grouping pattern of the genotypes between the genetic diversity and geographical origin of genotypes. Similar confirmations were also reported by the findings of Singh et al., (2007);  Dhama et al., (2009); Katiyar and Dixit (2009); Yadav et al., (2009); Devi et al., (2010); Shrivastava et al., (2012); Supe et al., (2013) and Kumar and Kumar (2016).
 

 
The average intra and inter-cluster D² values with their corresponding intra and inter-cluster distance are presented in Table 4. The inter-cluster distances were greater than intra-cluster distances, which indicated the presence of considerable amount of genetic diversity among the genotypes studied. The greater the magnitude of intra and inter cluster distance the higher the variability among the cluster and within the cluster and vice versa. The results are in concurrence with the findings of Kumar et al., (2006); Singh et al., (2007); Singh and Mishra (2008); Katiyar and Dixit (2009); Sen and De (2017). The least value of intra cluster distance was found in cluster II (D² = 1.827) indicating the presence of less heterogeneous genotypes grouped in this cluster. Whereas, maximum value of intra-cluster distance was observed in cluster I (D² = 2.543) revealing the existence of maximum differences among the genotypes falling in this cluster, followed by divergence (D² = 2.352) for cluster VI, cluster III with (D² = 2.340), cluster V (D² = 2.158) and cluster IV (D² = 2.041). Hence, selection within these clusters may be exercised based on the highest area of desirable traits. In any breeding programme where the nature of crosses is to be evaluated, choice of diverse parents is of paramount importance as they produce superior off-springs in the segregating generation than the closely related ones. The inter-cluster distance (D²) being the main criterion for selection of genotypes was also worked-out as crossing of genotypes within the same cluster would not produce superior off-springs. A range of 2.462 to 6.471 was observed when inter-cluster D² values were used to study the diversity among the clusters. The minimum value of inter-cluster distance (D² = 2.462) was found between cluster II and III indicating close relationship and similarity for most traits among the genotypes included in these clusters. Whereas, cluster I and IV showed maximum value of inter-cluster distance (D² = 6.471), followed by cluster IV and V (D² = 4.700) and cluster I and III (D² = 4.626) indicating that the genotypes included in these clusters are not so closely related showing good amount of diversity. Hence, these genetically diverse genotypes can be used as promising parents for hybridization. These results are corroborated with the findings of Kumar et al., (2007); Singh et al., (2007); Devi et al., (2010) and Shrivastava et al., (2012) as they also gave similar conclusion.
 

 
Diversity among the genotypes was also estimated based on the considerable amount of variation in cluster means for different character. Different clusters exhibited distinct mean values for almost all the sixteen characters which reflect the genetic differences between the clusters (Table 5). It is evident from the cluster mean table that the genotypes in cluster I had highest mean values for number of pods per plant, green pod yield per plant, shell weight per plant and seed yield per plant. Whereas, the genotypes of cluster IV showed the maximum mean for days to 50 per cent flowering, plant height, length of first fruiting node and number of seeds per pod. Comparative assessment of cluster means showed that for improving specific characters, the genotypes should be selected from the cluster having high mean value for that particular character. This comparison indicates that clusters I and IV had better cluster means for most of the characters, therefore, these clusters might be considered better for selecting genotypes as divergent parents. The similar results are exhibited with the findings of Kumar et al., (2006); Devi et al., (2010) and Shrivastava et al., (2012).
 

       
The principal component analysis (PCA) is one of a series of techniques for collecting high-dimensional information and using the dependence between the variables in a more tractable form without any loss of information. It represents the major contributor to the total difference in each differentiation axis. Based on the analysis the first five principal components having eigen values greater than one contributed 86.70 per cent of the total variability of 55 pea germplasm (Table 6 and Fig 1). Proportion of variance for the first 4 components were 35, 19.2, 14.4, 10.9 per cent respectively. PC-I showed positive association towards days to 50 per cent flowering, plant height, number of first fruiting node and length of first fruiting node. Whereas, PC-II positively associated with number of first fruiting node and width of the pod.  Similar findings were reported by Maqbool et al., (2010) and Baranwal et al., (2013). For first two principal components, explained variation among all the pea genotypes was graphically represented in scattered plot (Fig 4).
 

 

 

 
Partitioning clustering was also performed by K-means cluster analysis based Mahalanobis genetic distance. Nbclust: R package used to obtain optimum K value (K=4) for clustering 55 pea genotypes collected from the two different locations Fig 2. Clustering pattern pea genotypes using K means clustering were presented in Fig 3. Moreover, the dendrograph showing both hierarchical clustering and K-means (K=4) based partitioning clustering presented in Fig 5. This result was in accordance to Charrad et al., (2014).
 

 

 

Based on Mahalanobis D² analysis it can be concluded that the pea germplasm in the present study can be successfully used for planning future breeding programmes. The inter-crossing of genotypes showing the greater genetic divergence for most of the characters studied should result in superior heterotic crosses and also, generate valuable segregants in the later generations. It is expected that better performing varieties could be generated to increase productivity in field pea. Therefore, from the present study genotypes of cluster I and cluster IV  based on their high values for inter cluster distance, cluster means and K clustering can be hybridized as the potential parents to produce superior off-springs in the segregating generations and to improve pea productivity.