Genetic Diversity Studies of Soybean [Glycine max (L.) Merrill] Germplasm Accessions using Cluster and Principal Component Analysis

Soybean is a miracle golden bean of the 21st century because of extraordinary high protein content (38-43%) combined with high amount of oil (17-19%). Soybean was originated from north-east Asia, particularly China (Hymowitz and Newell, 1981). Soybean [Glycine max (L.) Merrill] is well-adapted to different agro-climatic conditions of tropical, sub-tropical and temperate zones. Soybean ranks first among oilseeds in the world and has now found a prominent place in India. World soybean production was estimated as 385.527 million tonnes. Brazil ranks first in soybean production with 144 million tonnes followed by the United States, Argentina, China and India (Anonymous, 2021). Production in India accounts for 12.04 million tonnes from11.45 million hectares with average productivity of 1051 kg/ha. Madhya Pradesh is the soybean bowl of India, contributing more than 89 per cent of the country’s soybean production, followed by Maharashtra and Rajasthan (Anonymous, 2022).

The genetic diversity analysis helps in selecting appropriate parents for combining new alleles for the trait in crop improvement programmes (Shadakshari et al., 2011). The principal component and cluster analysis have been successfully used to classify and measure the pattern of genetic diversity in soybean germplasm (Kayan and Adak, 2012). The PCA was carried out to determine the genetic relatedness of genotypes, the interdependence of different traits and the significance of traits in relation to total variance. The cluster analysis helps in grouping the genotypes based on the variation occurring among them. The genotypes selected from these two techniques could be used for selection.

During the rabi season of 2021-22, the experiment was conducted at the Department of Pulses, Tamil Nadu Agricultural University, Coimbatore. The experimental material consisted of 135 soybean genotypes including five check varieties viz., NRC 132, NRC 142, NRC 147, MACS 1460 and CO (Soy) 3 is given in (Table 1).



The experiment was set up in an augmented block design II with row to row and plant to plant distance of 30 cm x 10 cm respectively. To determine the genetic divergence across the genotypes, Ggt 2.0 software was used to construct the tree diagram and the clusters were formed using Euclidian distances and UPGMA tree clustering method. The components were extracted using the Principal Component Analysis approach by using the GRAPES software (Kujane et al., 2019).

Cluster analysis
 
The analysis of variance showed highly significant differences among the 135 soybean genotypes for all the ten quantitative traits studied. The 135 soybean genotypes were grouped into twelve clusters by using UPGMA tree clustering in Ggt 2.0 software as given in (Table 2). The cluster IV was the largest cluster with 30 genotypes followed by clusters VIII, II, V and III with 24, 22, 13 and 10 genotypes respectively. Cluster VI and VII were the solitary clusters with genotype viz., MAUS 60 and JS 98-61 respectively. The genotypes within the same cluster showed less variation whereas between the clusters exhibited maximum variation. Therefore, the genotypes from different clusters could be selected for crop improvement.


 
Principal component analysis
 
Principal component analysis was performed using the mean data of ten quantitative traits using GRAPES software. The quantitative traits of 135 germplasm accessions were categorized into ten different principal components based on the total variation. The Eigen value of first four principal components among ten PCs were more than one as given in (Table 3) and these four PCs contribute the cumulative percentage of variation of 79.77 per cent. The contribution of each trait to total variation is presented in (Table 4).





Reddy et al., (2021) reported 68.61% of variance was contributed by first three principal components out of ten principal components formed with 24 french bean germplasms. Kumar et al., (2010) observed 95% of total variation was contributed by first ten PCs out of the fourteen PCs formed with 64 groundnut breeding lines.

The first principal component showing 42.17% variation was associated mainly with number of clusters per plant, number of pods per plant and number of branches per plant. The outcomes of Ghiday and Sentayehu (2015) and Dubey et al. (2018) were similar with the present study for number of branches per plant and number of pods per plant. Jain et al., (2021) reported 28.6% of total variation was contributed PC1 and is associated with number of pods per plant, days to flowering and plant height in 40 chick pea genotypes. The second principal component contributing 15.72% variation was mainly related to hundred seed weight and single plant yield. Similar outcomes were reported by Singh and Shrestha (2019) and Dubey et al. (2018) for hundred seed weight and single plant yield respectively. The third principal component conferring 11.24% variation was mainly connected with number of pods per plant and number of pods per cluster. Similar findings were observed by Ghiday and Sentayehu (2015) for number of pods per plant. The fourth principal component exhibiting 10.64% of variation and it was mainly coupled with hundred seed weight and Dubey et al. (2018) and Singh and Shrestha (2019) had similar findings on hundred seed weight.

The scree plot given in (Fig 1) clearly depicts that PC1 had highest variation, followed by PC2, PC3 and PC4. Based on PC1 and PC2, the genotypes were scattered along the biplot as shown in (Fig 2).





The cos2 loading value ranges from 0.00 to 1.00 and these values were used to indicate the divergence among genotypes. MACS 1460, EC 18736 and PK 1038 were the genotypes far apart, while JS 89-24, NRC 25, NRC 2007-G-1-13, NRC 43 and PK 7247 were the genotypes closer to the origin. The genotypes away from origin with high cos2 loading value exhibited maximum divergence whereas genotypes closer to origin with less cos2 loading value exhibited minimum divergence. The quantitative trait contribution to total variation and its interrelationship is shown in (Fig 3).



The traits viz., number of clusters per plant, number of pods per plant and single plant yield were far away and contributes maximum variation while the traits viz., number of seeds per pod, number of pods per cluster and hundred seed weight were closer to the origin and showed minimum contribution to total variation. Similar finding was reported by Vijayakumar et al. (2022).

The traits number of clusters per plant, number of pods per plant and single plant yield and genotypes MACS 1460, EC 18736 and PK 1038 could be used for future crop improvement programme. In addition, the genotypes MAUS 60 and JS 98-61 were grouped in solitary clusters and therefore these genotypes could also be used for breeding work.

Authors are grateful to all professors of Department of Pulses, Tamil Nadu Agricultural University, Coimbatore for providing facilities and their immense support to complete the research work successfully.

None