To fulfil the food requirements of the growing population, plant breeding efforts and management practices have progressed the grain yield in important food crops like rice, wheat, and maize (Fischer and Edmeades, 2010) and finger millet (Adugna et al., 2011; Megha et al., 2023). Finger millet contributes to a lesser share of the country-wide food production in India, but it has significance in nutritional superiority, drought resilience, and regional food security, with a positive response to additional irrigation and fertilizers (Thilakaratna and Raizada, 2015). However, in the last two decades, finger millet productivity is decreasing (Adugna et al., 2011; Saxena et al., 2018; Megha et al., 2023). Therefore, to increase the grain yields further, donor lines with higher yield contributing traits are necessary (Nanja Reddy et al., 2021b), and one of the principal approaches for developing donor lines is the utilization of genetic resources (Upadhyaya et al., 2007). Utilization of distinct donors in hybridization for particular traits, developing the MAGIC population, and the selection of transgressive segregants would increase productivity further (Upadhyaya et al., 2007).
4.1 Genotypic variability for yield contributing parameters
In identification of donor lines for a specific character, the prerequisites are the availability of germplasm, genetic variability for specific traits, heritability, and stability over the locations/ seasons (Upadhyaya et al., 2007). In the present study, the genotypes were stable across the seasons for most of the traits including grain yield with lower F-ratios (statistic) than the F-critical value, and thus appropriate in the selection of donors (Table 1; Lule et al., 2012). In the combined analysis, the genotypes were stable for mean ear-head weight, and grain yield with no significant differences between the years. Hence, the genotypes have been utilized to identify superior donors over the popular variety, GPU-28 (Table 2; Suppl. Table 1). The existing wide genetic variability for other parameters was effective in selecting genotypes for specific superior traits. In particular, Acc. GE-4596 and GE-4683 had higher mean ear-head weight and cv. PES-110, PR-202, and RAU had higher ear-head numbers compared to cv. GPU-28. The popular variety, GPU-28, is a shy tillering variety with 2.5 tillers per hill with relatively a higher mean ear-head weight (Nanja Reddy et al., 2019). Hence, incorporating higher productive tillers of selected donors into the background of GPU-28 would increase finger millet productivity.
4.2 Contribution of specific traits to grain yield
Although genetic variability exists for yield-contributing traits and selection could be possible, the nature and extent of interrelationships between the characters will provide additional support in the selection of trait-specific donors (Akhtar et al., 2011). Pearson correlations revealed that the grain yield was positively and significantly correlated to both the dependent and independent traits. The correlations suggest that mean ear-head weight would be a better trait for yield improvement, as mean ear-head weight and ear-head number are strongly and negatively correlated (Table 3; Owere et al., 2015; Mujahid et al., 2020; Chaithra and Nanja Reddy, 2023). In addition, dividing the correlation influence of traits into direct and indirect effects by path analysis will be more meaningful in selecting yield-contributing characters and provides a better understanding of their association with grain yield (Dhavaleshvar et al., 2019). In the present study, among the independent traits, the mean ear-head weight followed by ear-head number showed a higher direct effect on grain yield with a lower residual effect (Table 4; Chaithra and Nanja Reddy, 2023). Furthermore, the mean ear-head weight, ear-head number, and threshing percentage collectively contributed to the grain yield by 86.7%, with a low residual value of 0.133 (Table 5; Nanja Reddy et al., 2021b), and hence, these traits are collectively appropriate for selecting the donors in finger millet, that can be used in development of MAGIC population.
4.3 Yield prediction by specific traits
The relationship between observed and predicted grain yield using MLR was statistically significant, suggesting that the model is appropriate. However, the grain yield of popular cultivars is already high over the germplasm accessions (Fig. 2). Therefore, direct selection for a given location or across locations directly for the grain yield is not advisable (Simion, 2020; Nanja Reddy et al., 2021b). Hence, trait-specific donor selection could be a better approach for improving the grain yield of finger millet, using the available genetic resources for independent traits such as mean ear-head weight, and productive tillers (Owere et al., 2015; Kandel et al., 2019). Furthermore, the theoretical model predicted a 29.5% and 17.8% increase in grain yield of cv. GPU-28 with incorporation of higher ear-head number, or mean ear-head weight of selected donors respectively (Table 7; Ojulong et al., 2017; Krishna et al., 2021; Nanja Reddy et al., 2021b; Nanja Reddy, 2023). The trait, higher productive tillers per plant could be better because the cv. GPU-28 has relatively a higher mean ear-head weight. However, both can be used in development of MAGIC population.