The normal distribution of salinity score, observed in the 50 rice accessions used in this experiment, is suggestive of the polygenic nature of the trait, thus confirming earlier findings. The present investigation revealed 17 SSR markers, and 43 alleles in the 50 genotypes. Among these markers, the number of alleles per locus varied between two and five, with an average of 2.7059 per locus. This low number of alleles per locus, indicates low diversity among genotypes, as has been reported earlier. The PIC values of the 50 rice accessions studied ranged from 0.2516 to 0.5350 with an average of 0.3665. Our study revealed that the primers RM483, RM562, and RM1287 show a number of alleles with PIC value of more than 0.5. This indicates the efficiency of these primers to detect heterogeneous accession, agreeing with the findings of Giarrocco. Though it can be argued that the use of a higher number of markers to characterize accession would be more effective in describing genotypes, our results show that a fewer number can be as efficient for identifying salt-tolerant genotypes. In the present study, out of 20 SSR markers, only RM8094 and RM3412 could discriminate the salt-tolerant genotypes from the susceptible genotypes. Analysis of genetic divergence in the 50 genotypes revealed the superiority of SSR markers over morphological traits, in elucidating genetic relatedness more precisely. Eleven clusters were obtained using SSR markers as compared to only eight clusters obtained using morphological traits. Similar comparisons of haplotypes for seedling stage salinity tolerance were done in previous studies. Babu et al. could delineate 14 haplotypes for six informative Saltol associated markers analyzed across 23 rice genotypes. Kordrostami et al. identified 14 haplotypes involving 12 Saltol associated markers across 44 rice genotypes, Krishnamurthy et al.; 2016b;analyzed 21 SSR markers across 94 rice genotypes and could identify 11 seedling stage salinity tolerant genotypes containing regions other than Saltol controlling their salinity tolerance.

As in these previous studies, we have found genotypes with probable novel allele regions which can be considered candidates for improving seedling stage salinity tolerance. To compare the presence of 12 key markers RM1287, RM483, RM10825, RM10720, RM3412, RM8046, RM8115, RM8094, RM6711, RM10871, RM5365 and RM10843 for salt tolerance, 12 haplotypes were identified among 50 genotypes, based on marker banding patterns. Forty genotypes had different combinations of CSR10 alleles at different loci, vertical grow table while ten genotypes did not share any allele  with CSR10. From the comparison of haplotypes with high frequency of CSR10 alleles it can be deduced that marker RM10843 showed association with high salt tolerance response. The marker RM10843 is present in salt-tolerant genotypes Kuzhiadichan, Poongar, and Sornamugi also. Some of the highly sensitive lines such as IR64, Soorkuruvai and Karunkuruvai also had alleles similar to CSR10 at this locus. Genotypes which carried alleles similar to CSR10 at marker loci RM6711, RM10871 and RM10843 showed differential reactions to salinity stress, which indicated that no single marker had a strong positive association with salt tolerance. The marker RM8046 helped distinguish salt-tolerant genotypes from a sensitive genotype IR64, which had alleles similar to CSR10. It is essential to validate gene-linked markers between donor and recurrent parent since it is used in marker-assisted backcross breeding. The other highly tolerant genotypes, namely Boomi and Garudan Samba did not possess any allele similar to CSR10 that could explain the tolerance, implying that they may possess novel QTLs alleles for salt tolerance. To conclude, the genotypes, Kuzhiadichan, Sornamugi and Poongar, possessed a high degree of salinity tolerance and, hence, can be used as new donors for the trait. In the other salt-tolerant genotypes, Boomi and Garudan Samba, the trait does not seem to be linked to saltol locus, and, therefore, they can become new sources for mapping QTL for seedling stage salinity tolerance. In future, these genotypes will be tested for their reproductive stage salinity tolerance.alinity is one of the major abiotic stresses directly affecting archipelago countries. Asia has the largest saline area  in the world and such areas continue to increase annually.

Salinity has negative impacts on rice production,posing a great challenge in increasing rice production in Asian countries including China, India, Indonesia, Thailand, Bangladesh and Vietnam, where rice is the primary crop. Thus, this problem requires attention and must be resolved. Salinity causes two major stresses, namely osmotic stress similar to drought  and ion toxicity stress especially Na+. High Na+ concentrations decrease soil osmotic pressure, limiting the ability of plants to absorb water from the soil. In addition, high Na+ concentrations decrease photosynthesis and often disturb ionic homeostasis equilibrium,especially over long periods, which may lead to plant mortality. Salinity stress in rice can be addressed using plant breeding programs. The use of doubled haploid  plants derived from anther culture is reportedly a fast and effective method to obtain rice lines tolerant to abiotic stresses. Safitri et al  produced DH rice lines putatively adaptive to salinity stress. Selection to obtain such adaptive lines can also be accomplished using an indirect approach. Therefore, the indirect approach can also be a solution to increase the effectiveness and efficiency of DH line selection. Indirect selection to salinity stress can be achieved using the following two approaches: field experiments to obtain rice lines with good agronomic characters and seedling screening in a greenhouse to obtain rice lines tolerant to salinity stress. However, both approaches require an effective and efficient selection method to determine the adaptive lines under salinity stress. One solution is to use an index selection method. This method involves a simultaneous selection with several characters formulated into a regression equation. The selection is based on the rank value of the regression index on each genotype based on the selection intensity. The advantage of this selection method is that each character is standardized and therefore they have the same data variance. It is more effective than single or direct selection. However, it is critical to determine the weight of each character in the selection process and multivariate analysis can be used to determine the weight of the selection index.

It can determine the important plant characters that can improve the effectiveness and efficiency of selection. In addition, it is also used to determine the value of weights in the selection index, such as path analysis,principal component analysis, factor analysis  and discriminant analysis  to determine the selection index formulas. Multivariate analysis is also expected to be used effectively and efficiently to select rice lines adaptive to salinity. This study aimed to determine the selection index formula and select DH rice lines adaptive to salinity using a multivariate analysis approach.Breeders are unable to study the interaction of genetic and environmental variance using only one location trial. Thus, selection might be ineffective if it only focuses on the yield, especially when the yield heritability is low. To increase the effectiveness, selection must include some secondary characters supporting the yield so that the selected entries can be relatively stable across environments. Besides, according to Acquaah  and Fellahi et al,secondary characters or the yield supporting characters have to possess a significant impact. Therefore, the determination of secondary characters can be conducted by employing the multivariate analysis approach, which can improve the accuracy of determination.However, these correlations had a low value, indicating that more specific multivariate analysis such as path analysis and multivariate regression should be used. Path analysis is identical to direct effects as the critical point to identify specifically related characters to the main character. Stepwise regression was the common regression used to set an ideal equation model. This method can reduce the bias of unimportant characters in the regression model, and the model can be more efficient in predicting a dependent variable. Therefore, the results from a combination of path analysis and multiple regression can be more effective and relevant than a general correlation to determine the selection characters. These combination methods in selecting the best secondary characters have been reported by Al-Sayed et al  for sugarcane, Saed-Mouchesh et al  for wheat, Kose et al  for spring safflowers and Anshori et al  for rice.

NPT and NFG were identified as ideal characters to support the yield because these two characters had the largest direct influence in the path analysis,and they played a role in predicting the yield in multivariate regression. These results also emphasized that one should avoid using character because it might be insufficient to explain a considerable proportion of the main character. The yield is the main character of selection so that the determination of representative PCs was in the direction with the greatest yield variance. Based on the yield variance, the eigenvectors in PC2 were considered as representative weight values for the selection indices. This finding was also supported by the direction of supporting yield characters, which were the same as the yield. A negative value only shows the absolute position of the character variance direction in a PC  so that the eigenvector value can be used as the basis to select character weighting. The use of PCA in developing a selection index has been reported by Godshalk and Timothy,Akbar et al,Alsabah et al  and Anshori et al. The multivariate analysis concept is considered effective and efficient because it can minimize errors or bias when determining the selected character and selection index formula. This was evident from an increasingly tight selection on GAI selection  compared to a direct selection. The use of check varieties can also improve selection accuracy because they are the product of breeders and have been tested for their characters, especially for their yield and yield components. As for the standardized yield, mobile vertical grow tables the selection of good agronomic lines was based on z-vDOXH• 0. This concept has been reported by Pertenelli et al  in clustering sugarcane genotypes. Therefore, the conceptual approach of GAI can be useful to minimize environmental effects and increase the selectivity on the DH line selection. Determination of the main character in response to salinity can be achieved using two approaches, namely the effect of genotype and environment  interactions  and their correlation to STSs. Significant interactions indicated that there were different character responses across genotypes under salinity stress. However, character responses should be significantly correlated with STSs as a general standard evaluation under salinity stress to distinguish tolerant and sensitive genotypes. In this step, the data had different levels: the STS value was ordinal, whereas the growth characters were numerical, and therefore Spearman correlation was used. 

The correlation results were also reported by Ali et al  who showed that plant height and biomass characters negatively correlate with salinity tolerance. STSs have a reverse orientation to the interpretation, with lower genotype scores indicating more tolerant genotypes. Therefore, the characters having a high negative correlation with STSs and significant G × E interaction served as the basis for determining the selection characters in salinity hydroponic screening. TFW was not included as a selection character. However, in both analyses, it showed a significant interaction and high correlation with STSs. TFW was strongly influenced by SFW. This can be seen from a non-significant interaction of root fresh weight and a significant correlation  between TFW and SFW. These results indicated that SFW was affected more than RFW under salinity. Therefore, the suitable characters to be used as selection characters in the screening of hydroponic culture salinity were SH and SFW. This finding was supported by a larger relative decrease in the shoot characters compared to other characters. In this study, the SaTI model in hydroponic culture was determined by combining the STS grouping and previously obtained selection characters. The scoring method was extensively used to screen salinity tolerance lines relying on the visualization of leaf chlorosis, which is a common symptom of salinity stress. However, the method was qualitative and did not consider the ability of plants to maintain growth rhythm under stress. Therefore, this combination was expected to improve accuracy in determining tolerance characters. The combination was developed through SSI based on discriminant function from the selection characters  and standardization of STS. The discriminant analysis produced a model which can linearly separate genotypes by group  and is widely used in the development of selection index. Here, it distinguished tolerant and sensitive line groups. The development of discriminant indices involved relative decreases in SH and SFW  because there were different responses among genotypes to environmental conditions for both characters. Therefore, the relative decrease was necessary to fairly consider genotypes.