3. Results and Discussion
The mean values for eleven characters under study estimated from the three locations were subjected to statistical analysis, location-wise and pooled. The mean values of parents, hybrids, and standard checks for the pooled data across locations (3) are illustrated using beanplots (
Figure 1). The mean value of parents was lower than crosses plus checks for all the characters except TGW. The difference between the means of the parents and crosses plus checks was very narrow for PH and SF. Depending upon the density of the data points, the shape of the bean plots changed for different characters under study.
The results from mean performance revealed that, among the lines, L02 was identified as good (considerably superior or on par with their respective mean) toward PL, PT, FG, BM, PPD, GY, and HI. L03 was considered good for DFF, PH, and TGW, while L01 was good for TGW. Testers T01, T02, and T07 recorded good GY, PT, TGW, and PPD. The testers, T03, exhibited earliness, and T10 recorded a short stature.
Substantially, depending on the inclusive performance, the following hybrids, H04, H18, and H20, performed in a superior way to the hybrid check, GK 5022, in response to GY, plus additional yield ascribing traits such as PL, PT, FG, SF, TGW, and PPD. The tables pertaining to the above results are furnished as
Supplementary Data for reference (Table S1).
The analysis of variance (ANOVA) for grain yield and yield ascribing traits unveiled significant differences among the genotypes (
Table 2) toward all the traits studied at every location. The significance of genotypes indicated the existence of commensurable variability amongst the tested genotypes.
Pooled ANOVA toward combining ability over locations unveiled significant differences amongst locations, genotypes (treatments), parents, parents vs. crosses, and crosses for all the traits studied (
Table S2).
The significance of parents, crosses, and parents vs. crosses for most traits studied has been previously reported by researchers [
17,
18]. The splitting up of crosses into components viz., lines, testers, and line × tester, also showed that variances were significant for traits studied. Furthermore, it witnessed significant variances for the line × tester component for all traits studied by rice workers [
17,
18]. The effect of the interaction of lines × testers × locations recorded substantial differences for the traits DFF, PT, FG, GY, PPD, and HI. Reports in agreement with the above findings presented significant variances of lines × testers × locations for PT, PL, FG, and GY [
17,
18].
These results expose the omnipresence of sizable variability within the plant material studied, and there is a reliable prospect for the identification of pragmatic hybrid combinations as well as parental lines.
The general combining ability (GCA) is linked with additive gene action, whereas the specific combining ability is traceable to dominance and epistasis. Pooled analysis unveiled greater SCA variances than GCA variances for all the traits, implying the preponderance of non-additive gene action, which was previously envisaged as ideal for exploiting full potential through heterosis breeding.
A comparative study of the measure of variance components due to
GCA and
SCA grounded the gene action nature in regulating the trait expression. The
GCA to
SCA variance ratio was less than unity, indicating the preponderant role of non-additive gene action for all traits studied, exhibiting a non-additive type of gene action (
Table 3). In support of present results, previous rice researchers documented findings envisaging the role of non-additive types of gene action for traits, namely DFF [
19,
20,
21,
22,
23], PH [
18,
20,
24,
25,
26,
27], PT [
20,
28,
29], PL [
17,
18,
26,
27,
30], FG [
18,
20,
24,
25,
27,
29], SF [
18,
25,
27,
30], BM [
22,
31,
32], HI [
31,
32,
33], TGW [
18,
27,
34,
35,
36], and GY [
18,
20,
23,
27,
30,
35,
36,
37,
38], as in the current experiment.
The contributory role of lines was recorded as high for four traits viz., PH, FG, PPD, and HI, while it was high for characters, i.e., DFF, PL, SF, TGW, BM, and GY (
Table 4). The line × tester interaction component contribution was higher for PT and modest for SF, with the characters being significant in deciding the hybrid potency, especially under aerobic conditions.
L02 was a good general combiner for PL, PT, FG, BM, PPD, HI, and GY, among lines. Out of ten testers, five were identified as excellent general combiners for GY as well as yield-attributing traits, including T02 for GY, PPD and HI; T04 for GY, DFF, PL, PT, FG, SF, BM, PPD, and HI; T06 for GY, PH, FG, and BM; T08 for GY, PL, PT, FG, SF, BM, PPD, and HI and T10 for GY, DFF, PH, SF, TGW, BM, and PPD (
Tables S3 and S4).
In a few cases, it was noticed that the lines and testers with good performance were not necessarily the best general combiners, and the opposite is also true. Thus, the choice of parents must be predicated on both (by itself) the expression and parent’s
gca effects. Line L02 was confirmed as a good combiner for GY and its ascribing traits. L02 has been previously reported as a good general combiner for GY [
39]. Amongst testers, T02, T04, T06, T08, and T10 were good combiners considering high
gca effects and for most of the yield ascribing traits. Hence, the above testers and lines are well-thought-out, potent donors for improving GY and linked components in upcoming rice breeding programs.
Among the crosses studied, 12 hybrids (H01, H03, H04, H07, H09, H15, H16, H18, H20, H21, H22, and H29) exhibited considerably positive
sca effects for GY. H27 (for DFF); H07, H10, H21, and H30 (for PH); H01, H08, H15, H27, and H28 (for PL); H04, H15, H18, and H20 (for PT); H04 and H20 (for FG); H04 (for SF); H07 and H11 (for TGW); H14 and H15 (for BM); H04, H18, and H20 (for PPD); and H18, H20, H22, and H24 (for HI) were identified as the best specific combiners based on considerable
sca effects (the above details in tables are furnished as
Supplementary Data for reference). However, H04, H18, and H20 (for GY) expression were exceedingly excellent for grain yield and its components regarding the good
sca effects of crosses and good
gca of parents. Here, it is clear that the significance of
sca effects alone has no effect as long as its mean value is in a desirable direction. Sometimes, the higher
sca effect may not be a choice among its counterparts after looking at the mean values. Hence, mean values have greater priority.
Thus, three outstanding specific combiners were detected amongst crosses, assumed from
sca effects and commensurable mean expression in descending order (
Tables S3 and S4). H20 for GY, PT, FG, BM, PPD, and HI; H18 for GY, PH, PT, BM, PPD, and HI; and H04 for GY, DFF, PL, PT, FG, BM, GY, PPD, and HI.
Heterosis toward grain yield/plant is predominantly because of concurrent exemplification of heterosis for the yield component character. Average heterosis or heterosis (h1), heterobeltiosis (h2), and standard heterosis (h3) arethe superior expressions as preferable over the mid parent, better parent, and the standard checks viz., GK 5022 (commercial hybrid) and CR–Dhan 201 (variety), projected for thirty hybrids for eleven traits (viz., DFF, PH, PL, PT, FG, SF, TGW, BM, GY, PPD, and HI for three locations and pooled data is a computed trait. The negative heterosis for DFF denotes earliness, and the negative heterosis for PH denotes short stature, which is preferable. In contrast, positive heterosis values were considered preferable for other traits.
The percentage of heterosis was calculated for pooled data pertaining to top specific-combiners for yield and yield-ascribing traits (
Table 5,
Tables S5 and S6).
As per the pooled analysis, average heterosis and heterobeltiosis estimates ranged from −42.29 (H11) to 131.13 (H04) percent and from −48.44 (H11) to 119.25 (H04) percent, respectively. Of the 30 hybrids studied, 18 excelled with considerable positive average heterosis and 16 exhibited considerable positive heterobeltiosis. Concerning heterosis, over best standard check GK 5022, the range was from −64.17 (H11) to 12.86 percent (H20) and positive significant standard heterosis was exhibited by four hybrids that included H20 (12.86), H18 (10.59), H04 (7.36), and H14 (3.20)
Heterosis and heterobeltiosis of the positive kind have been documented by previous workers in rice [
18,
27,
40,
41,
42,
43]. At the same time, few rice workers have proclaimed positive heterobeltiosis and standard heterosis values for this character [
18,
27,
40,
42,
43]. However, mean performance is also an important consideration coupled with
gca,
sca effects, and heterosis percentage [
44].
Further, top-ranking crosses were presented based on the high mean and their
sca effects, parent’s
gca effects, and standard heterosis for yield and its attributes (presented as
Supplementary Data). The hybrid, H20, which showcased extremely significant heterosis (positive) for grain yield compared to the checks, also proved its performance for PL, PT, FG, BM, PPD, and HI. Similar observations were noticed with H18 and H04 pertaining to GY and yield-ascribing traits. It was noticed in the cross combinations that involved lines IR-68897A and APMS-6A reported their superiority for GY [
39].
The stability ANOVA unveiled that genotypes and environments were significant for most traits except HI, signifying diversity amongst genotypes and environments (
Table S7). G × E interaction was considerable for the traits excluding PL, PT, TGW, and HI against pooled error, implying overwhelming behavioral differences of genotypes in erratic environments. G × E interaction for PL, PT, TGW, and HI were detected to be insignificant. Henceforth, stability assessment was not pursued for those traits.
Dissecting the sum of squares into varieties, environments + (genotypes × environment), and pooled error unveiled that mean squares owing to genotypes were highly considerable for all the traits examined, implying the manifestation of genetic variability in the studied experimental genotypic material [
18,
45]. Mean squares owing to environments + (genotypes × environments) were considerable for the entire range of traits except for TGW and HI. The above findings conformed to those of a few previous rice workers [
18,
45].
The sum of squares owing to environment + (genotype × environment) was further dissected into the environment (linear), genotype × environment (linear), and pooled deviation. Considerable variation owing to the environment (linear) was noticed for traits excluding HI, clarifying the linear contribution of environmental effects and additive environmental variance on these traits. Results in favor of the above findings have been documented by earlier researchers [
18,
45]. The linear component of G × E was considerable for traits excluding PL, PT, TGW, and HI, implying that genotypes considerably differ in their linear response to environments. The mean sum of squares for pooled deviation was considerable for DFF, PT, TGW, GY, PPD, and HI, implying the non-linear response and non-predictable nature of genotypes considerably differing in terms of stability. Thus, it unveils the significance of both linear and non-linear components in weighing the interaction of the genotypes with environments in the current study. The above findings conformed to those of a few previous rice workers [
18,
45,
46,
47].
As further stability analysis was not carried out for the following traits, viz., PL, PT, TGW, and HI, the adjudication of the promising experimental hybrids was made only considering their pooled mean expression.
Environmental indices of eleven characters viz., DFF, PH, PL, PT, FG, SF, TGW, BM, GY, PPD, and HI, are presented in
Table 6. The environmental index reveals how favorable one environment is at a peculiar location. It has been confirmed that the estimates of the environmental index can bestow the rationale for identifying the favorable environments for the expression of the maximum potential of the genotype [
15].
Environmental indices unveiled that Kampasagar was the most favorable location for FG, SF, TGW, BM, GY, PPD, and HI, while Warangal was the best location for PL and PT. Rajendranagar was the best location for DFF, PH, and PT.
Pooled ANOVA delineated the existence of considerable G × E interaction for GY. Linear and non-linear components pertaining to G × E interaction were considerable, which unveiled that only part of the performance could be predicted. A stable genotype, as per Eberhart and Russel (1966) [
14], exhibits (i) high mean yield, (ii) a regression coefficient (b
i = 1) equal to unity, and (iii) mean square deviation from regression (S
2di) near to zero. While comprehending the results of the current study, S
2di was considered toward the measure of stability, as suggested in [
15]. The estimates of stability parameters, i.e., mean (µ), the regression coefficient (b
i), and mean square deviation from regression (S
2di), were considered while assessing the stability of genotypes. The data related to stability parameters are furnished in
Supplementary Data for reference.
Among the genotypes, two lines, eight testers, twenty-one hybrids, and one check showcased inconsiderable deviations from the regression (S2di) values. Among the parents, one tester, T02 (20.54), exhibited average stability (mean significantly greater than varietal check, CR-Dhan 201) while another tester, T05 (13.49), was found to be adaptable to favorable environments (more than the average stability). None of the parents were found to be considerably superior over hybrid check GK 5022.
Two hybrids, H04 (32.78 g) and H20 (34.46 g), exemplified considerably higher GY over hybrid check GK 5022 (30.54 g) and recorded unit b
i values with non-significant deviation from regression. Hence, they were identified as highly adaptable hybrids and were thought to express well in various environments. H04 was also found to be stable for DFF, FG, and BM in addition to GY. Similarly, H20 was found to be highly adaptable for FG in addition to GY. Earlier rice researchers have also documented stable high-yielding GY hybrids based on stability parameters [
14,
46,
47,
48,
49].
Stable parents and crosses for grain yield and its component traits are listed (
Table 7 and
Table S8). Accordingly, parents, as well as crosses, are classified as stable and suitable to favorable environments and poor environments, respectively, based on the prescribed three features, i.e., mean (µ), the regression coefficient (b
i), and a mean square deviation from regression (S
2di).
Previous workers reported stable hybrids for various characters, viz., DFF, PH, and FG [
18,
45,
46] and SF [
18,
45].