Introduction
Genomic selection can improve the selection accuracy and reduce the generation interval. However, pedigree selection has been utilized for growth and body weight in commercial aquaculture breeding programs because of the increased cost of genotyping. Thus, in the present study an effort was made to develop a reduced, low-cost SNP panel to be utilized for genomic selection purposes for production traits (i.e., growth, body weight) and disease/parasite resistance (i.e., parasite count). The aim of the study is to compare the pedigree and genomic predictions using the 30K MedFISH array (Peñaloza et al., 2020) and selected low-density SNP panels for body weight in European seabass (Dicentrarchus labrax).
Materials and Methods
For this purpose, a selected sub-sample of 985 fish, from the total of 1,576 fish, infested with Lernanthropus kroyeri, was genotyped using the MedFISH array. Two GWAS were performed, one for growth in the sea cage and the other one for the final weight at the end of the experimental period (Oikonomou et al. 2022a, in press) and from these the significantly associated SNPs with the trait of interest (p-value<0.05) were selected. Further, a multitrait GWAS using different growth at different periods in the sea cage was performed and significantly associated SNPs (p-value<0.01) were selected. Based on the above criteria, 1,715 SNPs related to growth were finally selected.
Furthermore, a GWAS was performed for resistance to Lernanthropus kroyeri (parasite count, PC) (Oikonomou et al. 2022a, in press). Two low-density panels (SNP–panel 1 and SNP–panel 2) were constructed using two selected groups of SNPs as criteria, according to their p-value from the GWAS (p-value<0.01 and <0.05, respectively). Then, 245 and 1,192 SNPs associated to parasite resistance (GWAS for parasite resistance), were included in the SNP-panel 1 and SNP-panel 2, respectively. Τhose SNP panels were enriched with the 1,715 SNPs related to growth. Thus, two reduced density SNP – panels, which included 1,960 SNPs (named as SNP-panel 1) and 2,907 SNPs (named as SNP – panel 2) were constructed.
Estimated Breeding Values (EBVs) for the final body weight were calculated using Best Linear Unbiased Prediction (BLUP) and Genomic Estimated Breeding Values (GEBVs) for the final body weight were assessed using Genomic-BLUP (GBLUP) for the three SNP panels (30K MedFISH array, SNP–panel 1 & 2), using BLUPF90. Each time, 20% of the population was selected randomly and its phenotypes were masked (216 fish), thus, the breeding values (EBVs or GEBVs) were estimated using the information from 80% of the total fish. This process was performed 20 times and the correlation between the predicted values for the final body weight and phenotypes was calculated in the validation group. A one-way ANOVA with repeated measurements was performed among the four groups (SNP–panel 1, SNP–panel 2, 30K MedFISH array and pedigree).
Results
In the validation group, the use of pedigree showed the lowest correlation coefficient (0.38) followed by the 30K MedFISH array (0.41), while the use of SNP–panel 1 (1,960 SNPs) provided the highest prediction (0.70) followed by the SNP–panel 2 (2,907 SNPs) in which the estimate was 0.66 (Figure 1). Moreover, a statistically significant difference (p-value<0.01) among the four genetic evaluation procedures was found. The post-hoc analyses with a Bonferroni adjustment were performed and revealed that there was a statistically significant difference between each pair.
Discussion
When, selecting for disease resistance (parasite count) instead of body weight, the SNP – panel 2 provided the highest estimate (0.81) followed by the SNP – panel 1 (0.75) (Oikonomou et al. 2022b) while selecting for body weight the SNP-panel 1 provided the highest estimation (0.70). Regardless of the selection trait (disease resistance or body weight), higher correlation coefficient was established using small, selected SNP panels compared to the 30K MedFISH array.
Furthermore, higher correlation coefficient was found when using genomic information (30K MedFISH array and SNP-panel 1& 2) than when using pedigree relationship matrix. These findings indicate that small and cleverly selected SNP panels, such as SNP – panel 1 and 2 could potentially be utilized in cases of multi-trait genomic evaluation, and in some cases, can provide better predictions especially when parts of phenotypic information is missing.
Reference