Six-parameter model estimates

Simple additive-dominance model was inadequate, as revealed by significant values for A, B and C scales, suggesting the presence of non-allelic interactions in the genetic control of all the three sink size traits. These results were also supported by the three-parameter model of joint scaling test, showing large and significant χ² values. Therefore, six-parameter model was used to determine the type and magnitude of gene action involved in the inheritance of sink size traits.

4.1.1.4.1. Panicle length

For panicle length in cross 1, both additive and dominance effects were significant and negative in all the sets across seasons (Table 16). The additive effect (–17.3) was higher in magnitude than the dominance effect (–9.8) in set 3 during 2006 rainy season. This pattern was also observed in all other sets in both the seasons. The additive x additive (i) interaction was significant in most of the blocks and sets in both

the seasons, except blocks B₁ and B₃ in set 1, block B₁₃ in set 2 during 2006 rainy season and B₂ block of set 1 during 2007 summer season. Additive x dominance (j) and dominance x dominance (l) interactions were positive and significant in all the three sets in both the seasons. The magnitude of dominance x dominance interaction (25.5) was 81% and 68% higher than additive x additive (–4.72) and additive x dominance (8.06) interactions, respectively in set 3 during 2006 rainy season, and this pattern was also observed across other sets and seasons. The negative sign of dominance effect and positive sign of dominance x dominance interaction indicated a duplicate type of epistasis in all sets in both the seasons.

In cross 2, the average effect was highly significant for this trait in all the three sets in both the seasons (Table 17). The additive component was negative and highly significant in all the sets in both the seasons. Conversely, the dominance component was found to be non-significant across all sets in both the seasons. Among the interacting components, the additive x additive interaction was non-significant in set 1 across seasons, however in set 2, it was significant in all the three blocks during rainy season of 2006 and in block B₁₂ during summer season of 2007. Significance of this interaction was also detected in set 3 in both the seasons. The other interacting components, additive x additive and dominance x dominance effects were non-significant in all the sets across seasons. Both the dominance and dominance x dominance components were in the positive direction, indicating the presence of complementary type of epistasis, however, these components were non-significant in this cross.

4.1.1.4.2. Panicle diameter

The additive effect for this trait in cross 1 was highly significant in all the three sets in both the seasons (Table 18). The dominance effect was significant to highly

significant across the seasons in set 1, whereas in set 2 and set 3, this component was highly significant in both the seasons. The magnitude of additive effect (–15.9) was higher than the dominance effect (5.5) in set 3 during 2006 rainy season. Similar trend was observed across the other sets in both the seasons. The additive x additive interaction was significant only in B₁ block of set 1 during both the seasons. However, it was significant for all the block combinations in set 2 at varying levels in both the seasons. This interaction was also found to be highly significant in set 3 across seasons. The additive x dominance interaction was non-significant in all the sets in both the seasons except B₃ block of set 1 during 2007 summer season. However, the dominance x dominance interaction was highly significant and positive across all the sets in both the seasons, except B₃ block of set 1 during 2006 rainy season. The dominance x dominance interaction (10.8) was 65% higher in magnitude than the additive x additive interaction (3.8) in set 3 during 2006 rainy season. Similar trend was observed across all the sets in both the seasons. Complementary type of epistasis was inferred for this trait, as signs of both dominance (h) and dominance x dominance (l) component were in the positive direction.

In cross 2, the average effect was highly significant in all the sets in both the seasons (Table 19). The additive effect for this trait was highly significant and negative in all the sets in both the seasons. Significant and positive dominance effect was observed in block B1 of set 1 during 2006 rainy season and in all the three blocks of set 1 during 2007 summer season with varying levels of significance. However, this component was highly significant in set 2 and set 3 across all the blocks in both the seasons except in block B23 of set 2 during 2006 rainy season. The magnitude of additive effect (–13.3) was higher than dominance effect (3.9) in set 3 during 2006 rainy season. Similar trend was observed across other sets and seasons. The additive x

additive interaction was positive and significant in all sets across seasons. The additive x dominance interaction was non-significant in all the sets during rainy season of 2006, whereas it was significant in 2007 summer season for B₂ and B₃ blocks in set 1 and B₁₃ and B₂₃ blocks in set 2 with varying levels of significance.

However, it was detected with high level of significance in set 3. The dominance x dominance interaction was non-significant across sets and seasons. As dominance component was negative and dominance x dominance components was positive, duplicate type of epistasis was inferred for this trait in this cross.

4.1.1.4.3. Grain size

In cross 1, the additive effect for this trait was highly significant in all the sets in both seasons (Table 20). The dominance component was negative and significant in B1 and B2 blocks of set 1 during 2006 rainy season and B2 and B3 blocks during2007 summer season with varying level of significance. However in set 2 and 3, this component was highly significant in both the seasons. The magnitude of additive effect (–2.7) was higher than dominance effect (–1.5) in set 3 during 2006 rainy season. Similar trend was also observed across other sets and seasons. Additive x additive interaction was significant only in B1 block during 2006 rainy season and B2 block during 2007 summer season in set 1 at varying levels. Similarly, in set 2 also this interaction was significant in two blocks in both 2006 rainy season (B12 and B13) and 2007 summer season (B12 and B23). However in set 3, this component was highly significant across the seasons. Both the additive x dominance and dominance x dominance interaction were significant across all the sets and seasons. The dominance x dominance interaction (4.8) was 77% and 65% higher in magnitude than the additive x additive (–1.1) and additive x dominance (1.7) interaction components, respectively in set 3 of 2006 rainy season. This interacting component also showed a higher magnitude as

compared to all other components across all the sets and seasons. The negative sign of dominance component and positive sign of dominance x dominance component indicated the presence of duplicate epistasis in this cross.

The additive effect for grain size in cross 2 was highly significant in all the sets in both the seasons (Table 21). The dominance component was significant in set 1 for block B₂ and B₃ during 2006 rainy season and block B₂ during 2007 summer season. This component in set 2 was significant at varying levels across seasons.

However, it was highly significant in set 3 across seasons. The additive component (–

2.6) was 43% higher than the dominance component (–1.5) in set 3 of 2006 rainy season. This trend was also evidenced across all other sets and seasons. The additive x additive interaction was non-significant in all the sets in both the seasons except in set 3 during 2007 summer season. The additive x dominance interaction was highly significant in all the three sets during 2006 rainy season. During summer season of 2007, this component was significant at varying levels in set 1 and set 2 population sizes. However in set 3, it was highly significant across seasons. The dominance x dominance interaction was significant across all the three sets. The dominance x dominance interaction (3.46) was 51% higher in magnitude than the additive x dominance interaction (1.7) component in set 3 during 2006 rainy season. This interacting component was also higher as compared to all other components across all the sets and seasons. The negative sign of dominance component and positive sign of dominance x dominance interaction indicate a duplicate type of epistasis for this trait.

4.1.2.Triple test cross analysis

The detection, estimation and interpretation of epistasis has progressed much faster at the level of first degree statistics (Mather and Jinks, 1982) which has certain limitations due to the cancellation of genetic effects. The TTC technique of Kearsey

and Jinks (1968) tests the presence of epistasis and estimates additive (D) and dominance (H) components with a higher degree of precision in the absence of epistasis. Even in the presence of epistasis, it provides estimates of additive and dominance components, which are useful for comparison with variance estimates from filial generation data. In the present study, the detection of epistasis and additive and dominance components (assuming no epistasis) for panicle length, panicle diameter and grain size in one cross (cross 1) from each trait-specific group was carried as per the method given by Kearsey and Jinks (1968), and the results are presented in Table 22 to 24.

4.1.2.1. Panicle length

Analysis of variance showed that the interaction of blocks with additive x additive (i) as well as additive x dominance and dominance x dominance (j + l) epistatic component was non-significant. Therefore, the individual epistatic components were tested against total epistasis x block interactions. Total epistasis was highly significant for this trait. Further, partitioning of the total epistasis showed the highly significant contribution of additive x additive (i) and additive x dominance and dominance x dominance (j + l) interactions (Table 22). However, the relative magnitude of mean squares due to additive x dominance and dominance x dominance (j + l) interaction was higher as compared to additive x additive interaction.

On the assumption of absence of epistasis, analysis of variances for sums (L1i

+ L2i) and differences (L1i – L2i) revealed the significance of their respective mean squares. Accordingly, the additive effect and dominance component were highly significant for this trait. The relative magnitude of additive effect (371.6) was lower than the dominance component (465.1). However, as this cross gave evidence of significant epistasis for this trait, the estimates of the additive effect and dominance

component were biased by unknown extent. The average degree of dominance was in the range of overdominance, as the estimate was greater than unity (1.12). The correlation coefficient (r_sd) of sum (L_1i + L_2i) and differences (L_1i – L_2i) was negative and significant (–0.59) for this trait, indicating that dominant allele have increasing effects on the trait.

4.1.2.2. Panicle diameter

Analysis of variance showed that the interaction of blocks with additive x additive (i) as well as additive x dominance and dominance x dominance (j + l) epistatic component was non-significant (Table 23). Therefore, the individual epistatic components were tested against total epistasis x block interactions. The mean squares due to total epistasis was highly significant. The partitioning of the epistatic variance showed significant additive x additive and additive x dominance and dominance x dominance interactions for this trait.

The significance of mean squares due to sums and differences revealed the importance of both additive and dominance components. The estimate of additive effect (157.4) component was relatively higher than the dominance component (135.4). The degree of dominance was partial as evident from the estimate being less than unity (0.93) for this trait. The estimated value for correlation coefficient (rsd) of sum and differences was negative and significant (–0.67) for this trait, indicating that dominant alleles have increasing effects on the trait.

4.1.2.3. Grain size

Analysis of variance showed that interaction of blocks with additive x additive (i) as well as additive x dominance and dominance x dominance (j + l) epistatic component was non-significant. Therefore, the individual epistatic components were tested

against total epistasis x block interactions. Mean square for epistasis provided evidence for significant total epistatic effect (Table 24). When the overall epistasis was partitioned, the results showed non-significant additive x additive (i) epistasis for this trait. However, the additive x dominance and dominance x dominance (j + l) epistasis was highly significant.

Analysis of variance for sums and differences, on the assumption of no epistasis indicated significant mean squares for this trait. These results provide evidence for the presence of both additive and dominance genetic components for this trait. The estimated additive (9.87) component was lesser than the dominance component (12.34). The degree of dominance being more than unity (1.12) revealed overdominance for this trait. The estimated value for correlation coefficient (r_sd) of sums and differences was non-significant, indicating symmetrical distribution of dominant alleles among parents.

4.1.3. Estimation of components of variances and heritabilities

The variance components attributed to total genetic, additive and dominance variation along with dominance ratio and heritabilities (broad and narrow sense) were estimated for panicle length, panicle diameter and grain size. These estimates were obtained from the variance of six generations of trait-specific crosses evaluated during 2006 rainy and 2007 summer seasons. The results are presented in Table 25.

4.1.3.1. Panicle length

Panicle length in cross 1 had an additive variance (D) of 83.2 and 90.3 and dominance variance (H) of 14.2 and 9.0, while in cross 2 the additive variance was 43.0 and 55.8 and dominance variance was 4.9 and 3.2 during 2006 rainy season and 2007 summer season, respectively. The magnitude of additive variance was much greater than the

dominance variance, and the dominance ratio was less than unity in both the crosses and the seasons. Broad-sense heritability was high in both the crosses, ranging from 94.9 to 98.1% across seasons. Narrow-sense heritability ranged from 83.8 to 90.5%

across the crosses and the seasons.

4.1.3.2. Panicle diameter

Panicle diameter in cross 1 had an additive variance (D) of 31.0 and 29.9 and dominance variance (H) of 13.82 and 14.8, while in cross 2 the additive variance was 27.7 and 29.5 and dominance variance was –0.1 and –1.7 during 2006 rainy season and 2007 summer season, respectively. The magnitude of additive variance was higher than the dominance variance component, and the dominance ratio was less than unity in both the crosses across seasons. The estimated values for heritabilities in broad and narrow sense ranged between 90.5% and 95.9% and 64.0% and 91%, respectively across the crosses and the seasons.

4.1.3.3. Grain size

The additive variance for grain size was 9.1 and 5.2, while the dominance variance was –3.1 and –0.5 in cross 1 during 2006 rainy season and 2007 summer season, respectively. In cross 2, the additive variance was 5.2 and 7.1 and the dominance variance was –0.03 and –0.8 during 2006 rainy season and 2007 summer season, respectively.

The dominance ratio was found to be positive and less than unity in both the crosses across seasons. The estimated values for broad-sense heritability varied from 88.0 to 95.8% and narrow-sense heritability ranged between 89.0% and 96.1% across crosses and seasons for this trait.