Variability attributable to genes controlling target traits is referred to as genetic variation. As genes cannot express in vacuum and require appropriate environment such as spatial environment (location) in the present study, the degree of expression of traits controlled by genes is impacted by environment. Selection is effective only on variation attributable to genotype per se
and/or non-crossover genotype ´ environment interaction. It is therefore imperative to detect and quantify genetic and its interaction variation in breeding populations and/or germplasm accessions.
ANOVA is a diagnostic step to detect significance of different desired sources of variability for target quantitative traits. The accessions (genotypes), location and genotype × location interaction (GLI) contributed significantly to total phenotypic variability for all the traits except GLI for days to 50% flowering (Table 1).
Table 1: Pooled analysis of variance of horse gram germplasm accessions for five quantitative traits.
Significance of mean squares attributable to locations not only suggested differential expression of accessions in two locations, but also justified the selection of locations for evaluation of the accessions. However, substantially lower variance attributable to GLI compared to that attributable to genotype (Table 2).
Table 2: Estimates of components of phenotypic variability among horse gram germplasm accessions for five quantitative traits.
suggest comparable performance ranks of accessions in the two locations for all the five traits (Bernardo, 2020)
. BLUPs allow the comparison of accessions over time (generation, year) and space (location, block) by minimizing their effects (Bernardo, 2020
; Tajalifar and Rasooli, 2022)
. Hence, traits BLUPs were estimated within and across locations for all the germplasm accessions. In both the locations, the accessions varied widely as could be inferred from the estimates of absolute and standardized range and PCV for all the traits (Table 3; Fig 2).
Table 3: Descriptive statistics for five quantitative traits of horse gram germplasm accessions.
Fig 2: Box-Whisker and bar plots showing mean variation for five quantitative traits within and across locations.
Note: The length of bar plots represents the trait means, while total length of box and whisker line represents absolute range of traits.
The estimates of range provide clues about the occurrence of accessions with extreme trait expression. The estimates of unit-independent standardized range of the accessions were higher for primary branches plant-1
, pods plant-1
, pod yield plant-1
(g) and grain yield plant-1
(g) compared to that for days to 50% flowering, which were amply reflected by the estimates of PCV (Table 3). The magnitude of PCV varied substantially even among the former four traits. Similar results were reported by Priyanka et al., (2021)
and Visakh and Bindu (2022)
in horse gram. Given that GR evolve as a result of natural selection and/or human selection, substantial variability among the accessions could be attributed to accumulations of mutant alleles at loci controlling traits relevant to fitness (under natural selection pressure) and those relevant to domestication (Swarup et al., 2021).
Organization of variability
Hybridization is the most convenient and hence popular method of generating variability. Selection of the most suitable parents for use in hybridization is the key for generating useful and exploitable variability. Quantitative genetic theory suggests genetically diverse genotypes (though phenotypically similar) preferably those in which desirable alleles are dispersed between them are most likely to uncover high frequency of transgressive RILs. The rationale for maximizing the frequency of transgressive RILs is that plant breeding was/is being successful owing to the occurrence of transgressive RILS. Without the occurrence of transgressive RILs, plant breeding would not work (Mackay et al., 2021).
Such genetically diverse genotypes can be easily identified by classifying the genotypes into different clusters by maximizing the variability between clusters and minimizing variability within clusters. In the present study, we could classify the accessions into four clusters.
The quantitative trait mean differences among the four clusters were significant for all the traits (Table 4). The trait variances among the four clusters were also significant for all the traits, except for primary branches plant-1
Table 4: Estimates of quantitative traits means of horse gram germplasm accessions classified into different clusters.
Table 5: Estimates of quantitative traits variances among horse gram germplasm accessions classified into different clusters.
These results suggested effectiveness of ‘K-means’ clustering approach to classify the accessions into different clusters. Effective classification of accessions into different clusters enable reliable identification of genetically diverse genotypes for use as parents for generating breeding populations. The estimates of the mean of the five quantitative traits are contrasting among the accessions included in Clusters II and III. It is therefore desirable to choose accessions from among those included in Cluster II and III for various applications in horse gram breeding research and for developing improved cultivars. Through the increased use of these accessions, the efficiency of horse gram breeding for widely/specifically adapted and highly stable high-yielding pure-lines could be maximized. Several researchers classified varying number of horse gram accessions into different clusters. For example, Dasgupta et al., (2005)
classified 50 accessions into ten clusters, Singhal et al., (2010)
classified 88 accessions into five clusters, Geetha et al., (2011)
classified 100 accessions into sixteen clusters and Kanaka et al., (2015)
classified 38 accessions into two clusters respectively.
The use of trait-specific accessions help meet short-term objectives, as very often breeders are forced to meet immediate requirement of farmers, consumers and end-users. In the present study, several trait-specific accessions were identified (Table 6). A few of these accessions were desirable for a combinations of multiple traits (Table 7).
Table 6: Promising trait-specific accessions identified from horse gram germplasm collections.
Table 7: Estimates of quantitative traits means of the horse gram germplasm accessions promising for multiple traits.
While a few these accessions (IC 105785, Paiyur 1, Paiyur 2 and IC-139329) were comparable to checks, others (IC 139556, CRHG-19, NBPGRT-D179, IC 78605761 and Palem 2) were significantly better than both the checks (Table 7). Several previous researchers have also identified trait-specific accessions in different legumes. To quote a few, Kulkarni and Mogle (2011)
in horse gram, Meena and Kumar (2014)
in chickpea and Cobbinah et al., (2011)
in cowpea have identified trait-specific accessions.