Advances in Crop Science and Technology
Open Access

Lowland rice; Genetic diversity; Cluster analysis Principal component analysis; Hybridization; Segregation

Introduction

In Ethiopia, the cultivation of rice is of a recent history, however, its use as food crop, income source, employment opportunity and animal feed has been well recognized. The government of Ethiopia considered rice as the most strategic food security crop that has received special attention in promotion of agricultural production and as such it is named as the “millennium crop” expected to contribute in ensuring food security in The country expected to find export market in near future for other African countries and thus, the crop is one of the main trust areas of attracting for domestic consumption or international trade purposes.

Beside the country has huge potential for growing this crop, production, productivity and expansion of the rice has been hampered by lack of improved varieties, lack of recommended crop management practices for different rice ecosystems, lack of pre and post-harvest management technologies and lack of awareness on its utilization). As a result, rice yield remains progressively low with average national productivity of 2.8 tons/ha (CSA, 2015/16), which is very much lower than the average yield of rice in the world that accounts 4.54 tons per hectare according to FAOSTAT, 2015. Among the production constraints, lack of improved varieties is one of the most pressing constraint in rice production and productivity in the country.

Among the available options to tackle the low rice production and productivity in order to meet the fastest growing demand, exploiting genetic diversity which exists among rice genotypes for grain yield and yield components receives due emphasis. Diversity in plant genetic resources (PGR) provides opportunity for plant breeders to develop new and improved cultivars with desirable characteristics, which include both farmer-preferred traits (yield potential and large seed, etc.) and breeders preferred traits (pest and disease resistance and photosensitivity, etc.). One of the important approaches to rice breeding is hybridization and subsequent selection. Parents’ choice is the first step in plant breeding program through hybridization. In order to benefit transgressive segregation, genetic distance between parents is necessary (Joshi et al., 2004). The higher genetic distance between parents, the higher heterosis in progeny can be observed (Anand and Murrty, 1968). Estimation of genetic distance is one of appropriate tools for parental selection in rice hybridization programs. Appropriate selection of the parents is essential to be used in crossing nurseries to enhance the genetic recombination for potential yield increase (Islam, 2004). Some appropriate methods, cluster analysis and PCA for genetic diversity identification, parental selection, tracing the pathway to evolution of crops, center of origin and diversity, and study interaction between the environments are currently available. Principal component analysis helps researchers to distinguish significant relationship between traits. This is a multivariate analysis method that aims to explain the correlation between a large set of variables in terms of a small number of underlying independent factors. The cluster analysis is also an appropriate method for determining family relationships but the main advantage of using PCA over cluster analysis is that each genotype can be assigned to one group only. The main objective of this study is to assess the genetic diversity among rainfed lowland rice genotypes by using cluster and principal component analysis based methods for selection of parents in hybridization programme to obtain desirable segregants in advanced generation[1-4].

Materials and Methods

Description of the Experimental Site

The Trial was conducted during the main cropping season in July 2016 in two locations of Gimbo andYeki districts of Southwestern Ethiopia.

Description ofthe Experimental Materials

The experimental materials consisted of 25 rainfed lowland rice genotypes (Advanced lines) obtained from Fogera National Rice Research and Training Center (FNRRTC) and Bonga Agricultural Research Center Table 1.

Table 1: Description of the experimental materials

Source: Fogera National Rice Research and Training Center (FNRRTC) and Bonga Agricultural Research Center (BARC) Where, WARDA = West Africa Rice Development Association, IRRI = International Rice Research Institute., LRNVT = Lowland Rice National Variety Trial, LRRVT = Lowland Rice Regional Variety Trial

Experimental Design and Data Collection

The experiment was laid out in a 5x5 simple lattice design and five genotypes were assigned into each incomplete blocks. Standard evaluation system developed by IRRI13 was followed in order to collect yield and yield component data. Days to 50% heading and days to 85% maturity, biomass yield, harvest index, thousand seed weight and grain yield per hectare were computed from plot basis. Five representative plants for each genotype in each replication were randomly taken to record observations on plant height (cm), panicle length (cm), total tillers per plant, fertile or productive tillers per plant, filled grains per panicle, unfilled grains per panicle, and primary branches per panicle grain yield per plant. Grain yield obtained on plot base was converted into Kg ha-1 and adjusted to 14% grain moisture content[5-9].

Cluster analysis

Clustering was performed using the proc cluster procedure of SAS version 9.00 (SAS institute, 2002) by employing the method of average linkage clustering strategy of the observation. The number of cluster was determined by following the approach suggested by Copper and Milligan (1988) by looking into three statistics namely Pseudo F, Pseudo t2 and cubic clustering criteria. The points where local peaks of the CCC and pseudo F-statistic join with small values of the pseudo-t2 statistic followed by a larger pseudo-t2 for the next cluster combination was used to determine the number of clusters by using SAS statistical package (SAS Institute, 2002). The dendrogram was constructed based on the average linkage and Euclidean distance used as a measure of dissimilarity (the distance) technique.

The inter cluster distances were calculated by the formula described by Singh and Chaudhury (1997):

Square of the inter cluster distance = ΣD2i/ninj

Where, ΣD2i is the sum of distances between all possible combinations (ninj) of the genotypes included in the clusters under study, ni is number of genotypes in cluster i and nj is number of genotypes in cluster j.

Genetic divergence

Genetic divergence between clusters was calculated using the generalized Mahalanobis's D2 statistics (Mahalanobis’s, 1936) using the equation:

D²p = ((Xi – Xj) S–1(Xi- Xj).

Where, D2p = the squared distance between any two genotypes i and j; Xi and Xj = the p mean vectors of genotypes i and j, respectively. S-1 = the inverse of the pooled covariance matrix. The D2 values obtained for pairs of clusters were considered as the calculated values of Chi-square (X2) and tested for significance both at 1% and 5% probability levels against the tabulated value of X2 for 'P' degree of freedom, where P is the number of characters considered (Singh and Chaudhury, 1985).

Principal component analysis

Principal component analysis was performed using correlation matrix of SAS version 9.00 (SAS Institute, 2002) in order to examine the relationships among the quantitative characters that are correlated among each other by converting into uncorrelated characters called principal components. Below is the general formula to compute scores on the first component extracted (created) in a principal component analysis;

C1=b11(X1) +b12+…b1p (Xp)

Where, C1 = the subject’s score on principal component 1 (the first component extracted); b1p = the regression coefficient (or weight) for observed variable p, as used in creating principal component 1; Xp = the subject’s score on observed variable p.

Results and Discussion

Characteristics Cluster analysis grouped the genotypes into four distinct groups based on their similarity Figure 1. Genotypes obtained from Madagascar and WARDA were almost distributed in all clusters, indicating the existence of more genetic diversity in these origins and genotypes from the same origin might have different genetic background. The first cluster (C1, n=13) had the largest number of genotypes obtained from IRRI and Madagascar but, dominated by genotypes originated from Madagascar. The second cluster comprised eight genotypes, five of them were originated from Madagascar and the rest three were from WARDA. The third cluster consisted of three lowland rice genotypes, two of them were originated from WARDA and one from Madagascar and the fourth cluster only included one solitary genotype obtained from WARDA. The result showed that in most cases, genotypes tend to form their own distinct groups rather than forming their own hierarchical sub-groups based on their original background indicating seldom association between clustering pattern and eco-geographical distribution of genotypes, i.e., genotypes, collected from the same geographic area were placed into different cluster groups and those obtained from different geographic regions were placed into the same cluster (Figure 1).

Figure 1: Dendrogram indicating the genetic relationship of 25 rainfed lowland rice genotypes evaluated across two locations at south western Ethiopia in 2016/17 cropping season.

Numbers and corresponding genotypes are: 1. WAB189-B-B-B- HB, 2. IAC-164, 3. DEMOZE, 4. ROJOMENA271/10, 5. IRGA370- 38-1-1F-B1-1, 6. scrid113-3-5-3-5-4,7. scrid037-4-2-2-5-2, 8.scrid017-1-4-4-4-1,9. scrid014-1-1-1-1,10. scrid006-3-2-3-2, 11. Hibirre, 12. RPBIO4919-117, 13. IR82912-B-B-5, 14. IR88628- B-B-30, 15. IR83106-B-B-6, 16. scrid006-2-4-3-4-5, 17. WAB-4507, 18. FOFIFA172, 19. FOFIFA171, 20. FOFIFA161, 21. WAB502-8-5-1, 22 WABC165, 23 scrid079-1-5-4-2, 24. NERICA-18, 25. FOFIFA 165

Comparison of genotype performances among clusters

Cluster I was characterized by the highest cluster mean estimate for days to 50% heading, days to 85% maturity, plant height, number of filled grains per panicle and harvest index and it produced the lowest cluster mean estimate for grain yield hectare. Cluster II was characterized by having higher cluster mean values for number of total tiller per plant, number of fertile tiller per plant and biomass yield and the lowest cluster mean value for grain yield plant.

Eventhough cluster III displayed the second largest cluster mean estimates for several attributes viz., days to 50% heading, days to 85% maturity, number of primary branches per panicle, grain yield per plant, thousand seed weight and grain yield per hectare. It had the lowest cluster mean estimates for some of the traits evaluated viz., plant height, panicle length, number of total tiller per plant, fertile tiller number per plant, filled grains per panicle and unfilled grains per panicle. Cluster IV contained only one solitary genotype, i.e., WAB189-B-B-B-HB and it produced the highest cluster mean values for most of the traits i.e., panicle length, number of unfilled grains per panicle, number of primary branches per panicle, grain yield per plant, thousand seed weight and grain yield hectare and the lowest cluster mean values for days to 50% heading, days to 85% maturity and harvest index Table 2, indicating that this genotype is exceptionally the best genotype for performance under rainfed lowland condition.

Table 2: Cluster mean on 25 rainfed lowland rice genotypes for each of the evaluated traits

DH = days to heading, DM = days to maturity, PH = plant height, PL = panicle length, TTPP = number of total tiller per plant, FTPP = number of fertile tiller number per plant, FGPP = number of filled grains per panicle, UGPP = number of unfilled grains per panicle, PBPP = number of primary branches per panicle, GYPP = grain yield per plant, BY = biological yield, HI = harvest index, TSW = thousand seed weight, GY=grain yield per hectare

Distance among clusters (genetic divergence analysis)

Data The standardized Mahalanobis D2 statistics revealed the existence of high genetic distance among the four clusters and showed highly significant variation at P<0.01. The maximum squared distance was found between cluster one and four (D2=258.7) followed by cluster one and three (D2=191.5) and cluster two and four (D2=187.3). The minimum squared distance was found between cluster one and two (D2=45.06) followed by cluster three and four (D2=65.37) and cluster two and three (D2=124.6). Generally this study revealed that genotypes included in this study are highly divergent.

Minimum inter cluster distance was observed between cluster one and two (D2=45.06), indicating that genotypes in these clusters were not genetically diverse or there was little genetic diversity between these clusters. This signifies that crossing of genotypes from these two clusters might not give higher heterotic value in F1 and narrow range of variability in the segregating F2 population. Maximum genetic recombination is expected from the parents selected from divergent clusters groups. Therefore, maximum recombination and segregation of progenies is expected from crosses involving parents selected from cluster one and four followed by cluster one and three and cluster two and four Table 4.

** = significant, X2 = 27.69 and 22.36 at 1% and 5% probability level, respectively.

Principal component analysis

The Principal component analysis revealed that the first five components (PC1 to PC5) with eigenvalues greater than one explained 76.5% of the total variation among traits in rainfed lowland rice genotypes suggesting these principal component scores might be used to summarize the original 14 variables in any further analysis of the data (Table 3). Out of the total principal components retained, PC1, PC2 and PC3 with values of 26.7%, 17.0% and 14.3%, respectively contributed more to the total variation. The result is in agreement with the findings of by Robin et al. (2014) which identified five the most chief contributors in 192 rice germplasms for 12 agro- morphological traits that accounted for 80 % of the total variation. Mahendran et al. (2015) also identified the first 5 components that accounted for 77.38% of the total variation in rice.

Table 3: Pair-wise generalized square distance (D2) between four clusters constructed from 25 rainfed lowland rice genotypes

DH = days to heading, DM = days to maturity, PH = plant height, PL = panicle length, TTPP = total tiller number per plant, FTNPP= fertile tiller number per plant, FGPP= filled grains per panicle, UGPP= unfilled grains per panicle, PBPP = number of primary branches per panicle, GYPP = grain yield per plant, BY = biological yield, HI = harvest index, TSW = thousand seed weight, GY = grain yield per hectare.

The first principal component had high positive loading for 8 characters out of 14 viz. days to 85% maturity, panicle length, and number of total tiller per plant, number of fertile tiller per plant, filled grains per panicle, unfilled grains panicle, harvest index and thousand seed weight which contributed more to the variation. It has high negative weights for days to heading, biomass yield and grain yield per hectare. The major contributing traits for the variation in the second principal components (PC2) were chiefly obtained from variations of number of primary branches per panicle, number of total tiller per plant and harvest index and for the principal component three (PC3), grain yield per plant, biomass yield and number of total tiller per plant that contributed more to the variation. Days to 85% maturity and thousand seed weight expressed highest negative loads in principal component two (PCA2) The positive and negative weight shows the presence of positive and negative correlation trends between the components and the variables. Therefore, the above mentioned characters with high positive or negative loads contributed more to the diversity and they were the ones that most differentiated the clusters. A PC biplot in figure 2 showed that variables and genotypes are superimposed on the plot as vectors. The distance of each variable with respect to PC1 and PC2 showed the contribution of these variables in the variation of genotypes used.

Figure 2: Graphical representation (biplot) of the contribution of the first two principal components to the total variation in rainfed lowland rice genotypes.

According to Chahal and Gosal (2002), characters with largest absolute value closer to unity within the first principal component influence the clustering more than those with lower absolute value closer to zero. Thus, the prominent characters coming together in different principal components and contributing towards explaining the variability have the tendency to remain together. Hence, for the first group weight, number of fertile tiller per plant is best choice, which had the largest loading from component ones, thousand seed weight for the second, grain yield per plant for the third group and harvest index for the fourth group as the most chief contributors for the total variations in rainfed lowland rice genotypes. Therefore, in the present study, grouping of the genotypes into different clusters were because of relatively high contribution of these traits rather than small contribution from other traits. These results indicated the presence of genetic diversity and could be kept into consideration during hybridization and conservation programs [9].

Conclusions

Clustering and divergence analysis of quantitative characters based on multivariate analysis, indicated the existence four distinct groups and showed wide genetic diversity between different clusters. Thus, maximum recombination and segregation of progenies are expected from crosses involving parents selected from cluster one and four closely followed by cluster one and three and cluster two and four, respectively. Principal component analysis retained the first five chief contributors that played prominent role in explaining the total variation existing in the lowland rice genotypes. The analysis identified number of fertile tillers plant, thousand seed weight, grain yield per plant and harvest index in different principal components as the most prominent traits for differentiation of the total variation in lowland rice genotypes. Therefore, differentiation of genotypes into different groups was relatively because of larger contribution of these traits and might be considered in hybridization for evolving high yielding rainfed lowland rice genotypes. The present study indicated that there is adequate genetic diversity for most of yield and yield component traits evaluated in rainfed lowland rice genotypes for future exploitation and hence, could be kept into consideration during hybridization and conservation programs [9].

1. Salehi M, Vella A, McLaughlin T, Patti ME (2018) Hypoglycemia after gastric bypass surgery: Current concepts and controversies. J Clin Endocrinol Metab 103(8):2815-26.

Google Scholar, Crossref, Indexed at

2. Anand IJ, Murrty BR (1968) Genetic divergence and hybrid performance in linseed. Indian J Genet Pl Breed 28: 178-85.

Google Scholar 

3. Chahal GS, Gosal SS (2002) Principles and procedures of plant breeding: Biotechnological and conventional approaches.
4. Cooper MC, Milligan GW (1988) The effect of measurement error on determining the number of clusters in clusteranalysis. Data, Expert Knowledge and Decisions 319-328.

Google Scholar, Crossref

5. Eivazi AR, Naghavi MR, Hajheidari M, Pirseyedi SM, Ghaffari MR, et al. (2007) Assessing wheat (Triticum aestivum L.) genetic diversity using quality traits, amplified fragment length polymorphisms, simple sequence repeats and proteome analysis. Ann Appl Biol 152:81-91.

Google Scholar, Crossref

6. Hai lemichael G, Tesfaye K (2008) The effects of seed rhizome size on the growth, yield and economic return of ginger (Zingiber officinale Rosc.). Asian Journal of Plant Sciences 7(2): 213-217.

Google Scholar, Crossref

7. Hegde S, Hegde V (2013) Assessment of global rice production and export opportunity for economic development in Ethiopia. Int J Sci Res 2: 257-260.
8. Islam MR (2004) Genetic diversity in irrigated rice. Pak J Biol Sci 2: 226-29.
9. Joshi BK, Mudwari A, Bhatta MR, Ferrara GO (2004) Genetic diversity in Nepalese wheat cultivars based on agromorphological traits and coefficients of parentage. Nep Agric Res J 5: 7-17.