Transition from Conventional to Genomic Selection (ssGBLUP) led to improve in accuracy gains and selection decisions in Sahiwal Cattle Authors: Vikas Vohra, Parth Gaur, V. N. Sahana, Vikas Diwakar, Akshata Patil, Rani Alex, G. R. Gowane, T. V. Raja Published: 2026 ABSTRACT By using genome-wide markers to predict an individual’s genetic potential, the introduction of Genomic Selection (GS) has transformed animal breeding. This greatly accelerated selection for complex traits by lowering reliance on drawn-out field trials, allowing for faster genetic gains in livestock. However, there is little research on the effects of genomic selection on Sahiwal cattle in India, and comparing it to the current culling or selection process is even more uncommon, particularly in nations with fewer genotyped animals. This study is an initial effort to address the aforementioned gaps in knowledge. Genomic selection was implemented in Sahiwal cattle for the 305 days milk yield using univariate animal model and the single-step Genomic Best Linear Unbiased Prediction (ssGBLUP) method. The Effective Population size (Ne) of the Sahiwal herd was calculated using genomic data and was reported for the previous generation to be 71.927. The heritability of 305 days milk yield was estimated as 0.177 ± 0.068. Genomic estimated breeding values (GEBVs) were predicted for each individual using ssGBLUP, yielding a mean prediction accuracy of 43.11%, compared with 40.88% obtained using conventional pedigree-based BLUP. Cross-validation further demonstrated superior predictive performance of ssGBLUP, with accuracies of 76.82% and 70.50% for ssGBLUP and PBLUP, respectively. To further check the effectiveness of the genomic selection methodology, we also compared the GEBVs obtained and compared it with the Expected Progeny Difference (EPD) which is being applied in our farm for culling decisions. It was seen that GEBVs obtained from ssGBLUP methodology were also in line with the conventionally used method of EPD. The use of genomic selection enables genetic studies with limited pedigree information. Additionally, the ssGBLUP methodology allows to check for pedigree errors, where family relationships are incorrectly recorded. The EPD and GEBVs were consistent with one another, indicating that genomic selection may also be utilised to support culling and selection decisions in a farm. Thus, in a conventional animal breeding program with constraint resources and an incomplete pedigree, we recommend employing the ssGBLUP model for regular genomic assessment and identification of suitable candidates to effectively carry out a genomic selection program.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-63871-8. Introduction The goal of an animal breeder has always been to increase the genetic gain in the various economic traits in the farm animals, and for the aforementioned purpose, selection of superior farm animals is a pre-requisite. In the past, applying strict selection criteria has resulted in significant genetic gains in farm production and various selection criteria have been used over time1. In order to select an animal for higher productivity, the index-based approach used some dairy traits, and selection accuracy was established. However, the developed indices were found to be less robust to select superior animals based on a large number of traits because they required a continuous evaluation of farm economics and genetics over time, even within the same herd2. Later, the index-based strategy gave way to Best Linear Unbiased Prediction (BLUP), a mixed linear model methodology that used relative phenotypic data to forecast breeding values3. The breeding value (EBV) of dairy cows was estimated using BLUP-based selection, which produced accurate estimates of random effects and residuals. For economic features with strong estimates of heritability, BLUP proved quite successful. However, complex traits (dairy traits) with lower heritability estimates, those that are hard to measure, sex-limited, and those that manifest late in an animal’s life made BLUP-based selection challenging4. Beginning in the 1990s, advances in molecular genetics created the expectation that DNA-level information would accelerate genetic progress beyond what was achievable using phenotypic or pedigree data alone. Quantitative trait loci (QTL) mapping and genetic markers were proposed as tools for marker-assisted selection (MAS), with the goal of improving desirable traits by identifying genomic regions associated with variation5. However, for most livestock species, MAS had limited impact through the early and mid-2000s6. This was largely due to the difficulty of accurately identifying genes underlying complex quantitative traits, the high cost of generating suitable genotypic data, and the poor repeatability of QTL detection across studies. One explanation for this poor repeatability is that most QTL have relatively small effects. When combined with analyses involving large numbers of markers, this leads to the well-known “Beavis effect”, wherein estimated effect sizes of detected QTL are upwardly biased in studies with limited sample sizes7. Consequently, very dense marker panels and much larger datasets are required to reliably estimate QTL effects and maintain marker–QTL associations across families. The development of SNP genotyping arrays in the mid- to late 2000s provided access to high-density genomic information at relatively low cost. Nevertheless, even highly powered genome-wide association studies (GWAS) explained only a small proportion of the total genetic variance for most complex traits. This shortfall, termed the “missing heritability problem,” highlighted the inability of GWAS to capture the full range of heritable variation8. This realization emphasized that approaches focusing on individual loci were insufficient for predicting polygenic traits. To overcome these limitations, researchers provided a practical and highly effective solution to the limitations of MAS and GWAS, known as genomic selection9. Genomic selection estimates the genetic worth of the individual based on large set of marker information distributed across the whole genome10. In order to improve selection accuracy, maximise genetic gain, and lower costs, genomic evaluation techniques enable the direct use of data from an individual’s DNA in the genetic evaluation process11. The largest benefit of employing genomic evaluation is that it reduces the generation interval—the average age at which a sire becomes a parent—and increases genetic gains annually. Sires cannot be utilised to breed females in a breeding programme until their genetic merit has been established, which lengthens the generation interval. However, in a dairy herd, genomic breeding value (GEBV) can be calculated as early as the day of birth12. Therefore, the current study is an initial attempt to implement and assess the effectiveness of the genomic selection of Sahiwal cattle in India. Materials and methods Phenotypic data In order to conserve, improve genetically, and propagate Sahiwal cattle, the current study was conducted under the Institute project, ICAR NDRI, Karnal. The history cum pedigree sheet and performance register of Sahiwal cattle, which is kept in the Animal Genetics and Breeding Division’s record room at the ICAR-National Dairy Research Institute (NDRI), Karnal, provided the relevant data for this study. Between 1957 and 2024, a total of 1901 animals were examined in this investigation. 1435 animals’ first lactation records were extracted from the data repository of ICAR-NDRI. The sires in the pedigree were 148 and the total number of dams in the pedigree were 916. The season of birth (SOB), season of calving (SOC) and age at first calving (AFC) were also considered in the analysis for the correction of the fixed effect in the data. The year of birth was further converted into period of birth of birth by using the Sturges’ formula14 and seasons were defined as Summer from April to June, Rainy from July to September, Autumn from October to November and Winter from December to March, respectively. For checking the effect of each non-genetic factor least square analysis was performed15. The animals only above 500 kg were taken in the analysis so as to avoid the outliers in the data and thus making the data more normally distributed for the analysis. Genotypic data The genotyped animals were selected from the ICAR-NDRI Sahiwal herd, which represents the largest organized Sahiwal breeding population in India and one of the few herds with long-term systematic pedigree and performance recording. Over the course of this breeding program, pedigree and performance records have been maintained for approximately 5,130 animals. However, owing to selection practices and the requirement for complete and standardized first-lactation milk yield records, only about 1,500 purebred Sahiwal cattles were retained in the breeding and genetic evaluation program. From this breeding population, 501 Sahiwal cattle presently standing in the Institute herd (approximately 33%) with complete phenotypic records were selected for genotyping and subsequent genomic analyses. Under veterinary protocol, 10 ml of venous blood was drawn from each animal’s jugular vein and placed in a 15 ml polypropylene centrifuge tube with 0.5 ml of Ethylene-Diamine-Tetra-Acetic Acid (EDTA) acting as an anticoagulant after the IAEC (Institutional Animal Ethics Committee) approval for the same. Using the phenol extraction method, genomic DNA was separated, and the quality of the DNA was evaluated in a 0.08% w/v solution of Agarose in 0.5X Tris-borate-EDTA (TBE) buffer (pH 8), as described by16. Using the BovineSNP50 v3 BeadChip (Illumina Inc., San Diego, CA, USA), which analyses 53,218 SNP spread throughout the animals’ whole genome, SNP genotyping was carried out on 501 Sahiwal cattle. The PLINK programme v1.90 was used to undertake the quality control of the genotyping data17. Using the PLINK programme v1.90, recoding from A/B to 0/1/2 genotypic notation was also carried out. Statistical analysis The single-step Genomic Restricted Maximum Likelihood (ssGREML) technique was used to estimate the (co)variance component and genetic parameter for the 305 days milk yield, whereas the ssGBLUP method was used to predict the breeding values. BLUP was also performed to estimate the breeding value using pedigree data only (PBLUP) and it was used as baseline model to compare the breeding value obtained from ssGBLUP. The AIREMLF90 programme was used to conduct univariate analysis in an animal model18. The univariate animal model is described below: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathrm{Y}}_{{\mathrm{MY}}}}\,=\,\mu \,+\,{\mathrm{Xb}}\,+\,{\mathrm{Zu}}\,+\,{\mathrm{e}}$$\end{document} Where, YMY= Vector of observations of milk yield. µ = Overall mean. X= Incidence matrix relating b to YMY. b= Vector of fixed effects (YOB, SOB, SOC, AFC). Z= Incidence matrix relating u to YMY. u= Vector of random additive genetic effect. e= Vector of random residuals. The above model was selected out of four different models. The first model considered year of birth (YOB), season of birth (SOB), season of calving (SOC) as fixed effect and Age at first calving (AFC) as covariate; in second model AFC was taken as fixed effect and its classes were made using Sturges’ formula keeping rest parameter same; in third model instead of year of birth, year of calving classes were taken and fourth and last model is as described above. The last model was further used as it gave the lowest standard error in the estimation of variance and also had lowest AIC values. In the single-step approach, the inverse of the relationship matrix A is replaced by the inverse of relationship matrix H, that includes both pedigree-based relationships and differences between pedigree-based and genomic-based relationships19 as follows: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left[ {\begin{array}{*{20}{c}} {{\mathbf{X^{\prime}X}}}&{{\mathbf{X^{\prime}W}}} \\ {{\mathbf{W^{\prime}X}}}&{{\mathbf{W^{\prime}W}}+{{\mathbf{H}}^{ - 1}}\lambda } \end{array}} \right]\left[ {\begin{array}{*{20}{c}} {{\mathbf{\hat {b}}}} \\ {{\mathbf{\hat {u}}}} \end{array}} \right]=\left[ {\begin{array}{*{20}{c}} {{\mathbf{X^{\prime}}}y} \\ {{\mathbf{W^{\prime}}}y} \end{array}} \right]$$\end{document} Where, X = Incidence matrix of fixed effects. b = Vector of fixed effects. W = Incidence matrix of random effects. u = Vector of random effects. y = Observation of phenotypes. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda = \frac{{1 - h^{2} }}{{h^{2} }}$$\end{document} H = Augmented Pedigree–Genomic relationship matrix. and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\boldsymbol{H}}^{ - 1}}=~{{\boldsymbol{A}}^{ - 1}}+~\left[ {\begin{array}{*{20}{c}} 0&0 \\ 0&{{{\mathbf{G}}^{ - 1}} - {\boldsymbol{A}}_{{22}}^{{ - 1}}} \end{array}} \right]$$\end{document} in which A is the matrix of numerator of Wright’s coefficient of relationship; A22 is the section of A related to the genotyped animals; and G is the genomic relationship matrix, as proposed by the “first method” by VanRaden20. The prediction error variance was used to derive accuracy of breeding values21. In the mixed model equations, the PEV is obtained from diagonal elements of the inverse of the left-hand side within the inverse of the coefficient matrix22. The expected accuracy of GEBV for the ith individual was calculated from the prediction error variance (PEV) of the breeding value as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{Accurac}}{{\mathrm{y}}_{\mathrm{i}}}=\sqrt {1 - \frac{{PE{V_i}}}{{\sigma _{u}^{2}}}}$$\end{document} Where PEV is the prediction error variance and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sigma _{u}^{2}$$\end{document} is the additive genetic variance of the population. Further, the cross-validation methodology was also utilized to estimate the accuracy of the breeding value of the population. In this method, data was divided into random 5-fold, and each fold was used as testing dataset consecutively by estimating EBVs from remaining data23. EBVs were also estimated from full dataset that includes all animals, and EBVs of animals in each test dataset were extracted. Then accuracy estimates were calculated by correlating EBVs estimated from full dataset with EBVs estimated from each fold for the animals in the test set and average accuracy and slope estimates were calculated. The number of animals in training dataset were 1148 and in validation dataset were 287 in each fold. These animals were randomly selected in each set using syntax set.seed(123) and sample in the R 4.3.0 software. The effective population size (Ne) was also estimated using the software program SNeP1.11 and GONE24. The effective population size (Ne) were calculated using following parameters -ld (To save a .ld file with the distance and linkage disequilibrium value for each valid SNP comparison), -numBINS 500 (To define the number of bins in the .NeAll file), -expB 2 (To define the distance distribution for bins in the .NeAll file), -mindist 100,000 (Defines minimum distance in bp between SNPs to be analysed), -maxdist 100,000,000 (Defines maximum distance in bp between SNPs to be analysed), -svedf (To utilizes recombination rate modifier as defined by Sved & Feldman25. Expected Progeny Difference (EPD) for maternal milk were also estimated for each sire individually and was compared with the Genomic Estimated Breeding values (GEBVs) calculated earlier using ssGBLUP methodology. Results A basic data structure including simple arithmetic mean with standard error, standard deviation, and phenotypic standard deviation of 305 days milk yield for Sahiwal cattle along with preliminary pedigree information are presented in Table 1. After the standardization of data 1012 phenotypes were retained in the data and rest of animals were still included in the analysis for increasing the depth of the Numerator Relationship Matrix (NRM) to ensure the increased accuracy26. After standardization of the data, the average 305 days lactation was recorded as 1949.18 ± 26.59 kg. Table 1Basic data structure of 305 days milk yield in Sahiwal cattle.TraitNumber of observationsTotal number of animals1901Total number of sires148Total number of dams916Maximum paternal family size56Maximum maternal family size7Mean ± SE1949.180 ± 26.590 kgPhenotypic standard deviation848.095 Table 1Basic data structure of 305 days milk yield in Sahiwal cattle.TraitNumber of observationsTotal number of animals1901Total number of sires148Total number of dams916Maximum paternal family size56Maximum maternal family size7Mean ± SE1949.180 ± 26.590 kgPhenotypic standard deviation848.095 Basic data structure of 305 days milk yield in Sahiwal cattle. The study included the genetic dataset of sizeable number of 501 purebred Sahiwal cattle from the ICAR-NDRI, Karnal (Largest organised dairy herd of Sahiwal cattle in India). To guarantee consistency and dependability for genetic analysis, the genotypic data underwent quality control methods. PLINK 1.9 was used to eliminate mitochondrial SNPs, sex chromosomes, and unmapped SNPs from the dataset27. Genotypic call rates less than 90% (< 90%) for SNPs were maintained to reduce the amount of missing data and allow a more complete picture of the genetic information. To prevent uninformative SNPs, the minor allele frequency (MAF < 5%) was required. Additionally, monomorphic SNPs were eliminated, as were animals with a call rate less than 0.90. Following quality control of the genomic data, 501 animals’ genotypes on the 22,020 SNP were kept. Non-genetic effects were estimated using least squares analysis of variance (Table 2). It was seen that the milk yield of Sahiwal cattle was greatly impacted by the time of birth and age at first calving whereas season of calving and season of birth were found to have non-significant influence on the 305 days milk yield. Table 2Least squares mean for milk in Sahiwal cattle.SourceSum of squaresdfMean squareFSig.Intercept29493041.260129493041.2636.480.01POB51734173.470105173417.3476.3990.01*SOB860157.4423286719.1470.3550.786SOC5026609.47231675536.4912.0730.102AFC8140504.45918140504.45910.0690.002**Significant (p < 0.05), POB = Period of Birth, SOB= Season of Birth, SOC= Season of Calving & AFC = Age at First Calving. Table 2Least squares mean for milk in Sahiwal cattle.SourceSum of squaresdfMean squareFSig.Intercept29493041.260129493041.2636.480.01POB51734173.470105173417.3476.3990.01*SOB860157.4423286719.1470.3550.786SOC5026609.47231675536.4912.0730.102AFC8140504.45918140504.45910.0690.002**Significant (p < 0.05), POB = Period of Birth, SOB= Season of Birth, SOC= Season of Calving & AFC = Age at First Calving. Least squares mean for milk in Sahiwal cattle. *Significant (p < 0.05), POB = Period of Birth, SOB= Season of Birth, SOC= Season of Calving & AFC = Age at First Calving. During analysis we further encountered “Parent-Progeny Mendelian Conflict”. A “parent-progeny Mendelian conflict” occurs when the patterns of genetic inheritance found in a family’s genetic data do not match those predicted by Mendel’s Laws of Inheritance. This disagreement results from real-world problems including mis-genotyped individuals, where genetic markers are wrongly determined; pedigree errors, where family relationships are incorrectly documented; and the existence of unidentified parents28. In simple terms, the parent-progeny Mendelian conflict arise when there is mismatch between the pedigree and genotypic data, either due to incorrect documentation or some relationship between the previously asuumed unrelated animals. For genetic data quality control and to guarantee the validity of genetic research, it is essential to recognize these conflicts. After making the necessary corrections, we conducted a comprehensive analysis. The (co)variance component and genetic parameter for the 305 days milk yield were estimated by the single-step Genomic Restricted Maximum Likelihood (ssGREML) method by using the Blupf90 group of software and the model converged after 14 rounds of iteration and − 2logL value and Akaike Information Criterion (AIC) value of model were 16061.165 and 16065.165, respectively. Mean additive genetic variance was found to be 122,780 and mean residual variance was found to be 568,840 as shown in the Table 3. Table 3Variance and heritability estimate for 305 days milk yield.ParameterPBLUPssGBLUPVg88,897 ± 46,371122,780 ± 48,391Ve599,620 ± 47,983568,840 ± 47,110h20.129 ± 0.0670.176 ± 0.068Vg—additive Genetic Variance; Ve—environmental Variance; h2—heritability. Table 3Variance and heritability estimate for 305 days milk yield.ParameterPBLUPssGBLUPVg88,897 ± 46,371122,780 ± 48,391Ve599,620 ± 47,983568,840 ± 47,110h20.129 ± 0.0670.176 ± 0.068Vg—additive Genetic Variance; Ve—environmental Variance; h2—heritability. Variance and heritability estimate for 305 days milk yield. Vg—additive Genetic Variance; Ve—environmental Variance; h2—heritability. The heritability of the trait was found to be 0.177 ± 0.068. Further, the genomic estimated breeding values (GEBVs) were estimated for each individual and mean accuracy of the GEBVs was 43.11%. The heritability of the trait was found to be 0.177 ± 0.068. Further, the genomic estimated breeding values (GEBVs) were estimated for each individual and the mean prediction error variance (PEV)-based accuracy of GEBVs was 43.11%. Compared with PBLUP, which yielded a mean accuracy of 40.88%, the incorporation of genomic information through ssGBLUP resulted in an improvement in breeding value prediction accuracy. Furthermore, predictive performance was independently assessed using cross-validation. The cross-validation accuracies obtained using PBLUP and ssGBLUP were 70.50% and 76.82%, respectively, indicating superior predictive ability of ssGBLUP. The average slope in cross validation was 1.002 ± 0.0147, thus indicating less prediction bias. The effective population size (Ne) were calculated using parameters mentioned in the materials and methods section and the results are shown in the Fig. 1. Ne for the previous generation was found to be 71.927. Fig. 1The effective population size (in y axis) plotted against the ancestral generations (in x axis). The effective population size (in y axis) plotted against the ancestral generations (in x axis). Based on the genomic estimated breeding values (GEBVs) of the bull for the milk production, the selection and culling decision for the bulls were taken. A bull with a higher GEBV indicates that its daughter has a greater genetic potential for producing milk than an animal with a lower GEBV. Further, the culling decisions earlier were taken on the basis of the Expected Progeny Difference (EPD) in the herd so the EPD was also compared with the GEBV of the animals. The various animals and their GEBVs has been shown in the Tables 4 and 5. The animals with the higher GEBVs had also the higher EPD thus suggesting the GEBV can also be used as an aid for the selection and culling decisions to be undertaken at the farm. Table 4Top 10 selected bulls in the herd.Bull No/ setDOBDMY (kg)Genotype IDGenomic estimate (kg)RankEPD2979 M10-08-20223393SWM_2979-15.GType+ 300.284ProvenBull (1)1982840 M13-06-20202782SWB-2840.GType+ 289.50921642736 M28-11-20183502SWB-2736.GType+ 281.81431802683 M31-01-20182782SWB-2683.GType+ 277.11441682655 M/506-05-20173700SWB-2655.GType+ 267.32652682620 M09-02-20163459SWB-2620.GType+ 252.47161272509 M/508-05-20153702SWB_2509-10.GType+ 231.06372062888 M16-03-20213502SWB-2888.GType+ 204.04481352605 M/526-09-20163406SWB-2605.GType+ 198.19491973021 M20-04-20233760SWB-3021.GType+ 194.36210109DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Table 4Top 10 selected bulls in the herd.Bull No/ setDOBDMY (kg)Genotype IDGenomic estimate (kg)RankEPD2979 M10-08-20223393SWM_2979-15.GType+ 300.284ProvenBull (1)1982840 M13-06-20202782SWB-2840.GType+ 289.50921642736 M28-11-20183502SWB-2736.GType+ 281.81431802683 M31-01-20182782SWB-2683.GType+ 277.11441682655 M/506-05-20173700SWB-2655.GType+ 267.32652682620 M09-02-20163459SWB-2620.GType+ 252.47161272509 M/508-05-20153702SWB_2509-10.GType+ 231.06372062888 M16-03-20213502SWB-2888.GType+ 204.04481352605 M/526-09-20163406SWB-2605.GType+ 198.19491973021 M20-04-20233760SWB-3021.GType+ 194.36210109DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Top 10 selected bulls in the herd. Proven Bull (1) DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Table 5Culled animals which were genotyped and their corresponding EPDs.Bull No/ SetDOBDMY(kg)Genotype IDGenomic Estimate (kg)EPD2497 M08-03-20153208SWB_2497-7.GType-111.606752684 M01-10-20183591SWB-2684.GType-84.851-362895 M29-04-20212204SWB-2895.GType-91.252-152993 M02-12-20222540SWB-2993.GType-19.09522459 M22-07-20143455SWB_2459-13.GType-156.577-243018 M15-04-2023--SWB-3018.GType-124.31681DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Table 5Culled animals which were genotyped and their corresponding EPDs.Bull No/ SetDOBDMY(kg)Genotype IDGenomic Estimate (kg)EPD2497 M08-03-20153208SWB_2497-7.GType-111.606752684 M01-10-20183591SWB-2684.GType-84.851-362895 M29-04-20212204SWB-2895.GType-91.252-152993 M02-12-20222540SWB-2993.GType-19.09522459 M22-07-20143455SWB_2459-13.GType-156.577-243018 M15-04-2023--SWB-3018.GType-124.31681DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Culled animals which were genotyped and their corresponding EPDs. DOB—Date of Birth; DMY—Dam’s Milk Yield; EPD—Expected Progeny Difference. Discussion Average first lactation milk in the study (1949.18 ± 26.59 kg) was much higher than the national average for indigenous breeds (3.86 Kg/day/Animal) which make it 1177.3 kg. Further, these animals recorded higher average first lactation milk than the Pakistan Sahiwal (from 1424.23 ± 26.74 kg29 to 1552.1 ± 12.1530 which is considered the major home tract of this breed. However, the previous studies in NDRI gave the similar results (ranging from 1900.42 ± 93.3931 to 1939 ± 6932. This highlights the superiority of the elite herd of Sahiwal maintained in NDRI and suggests that if efforts are done towards further improvement of this herd, it can lead to increase in the national milk average from the indigenous animals. The first commercial SNP genotyping chip, the Illumina BovineSNP50 BeadChip containing 54,001 SNPs selected using a spacing algorithm, was introduced in 2007. This development, carried out as part of the Bovine Genome Project launched in 200333, provided evenly distributed markers across the genome at lower cost and enabled large-scale genotyping. Owing to which, 360,000 animals have been genotyped in France alone, whereas 934,780 Holstein, 120,439 Jersey, 19,588 Brown Swiss, and 4,767 Ayrshire animals have been genotyped in the USA alone1. Australia released its first formal genomic evaluations in 2011, and New Zealand started using the BovineSNP50 chip for its evaluations in 200834. In 2009, Canada published its official genomic evaluations after working with USDA to generate them35. Five European breeder-owned businesses (representing Belgium, Denmark, Finland, France, Germany, the Netherlands, and Sweden) founded Euro Genomics in 2009 with the goal of increasing evaluation accuracy in each nation by pooling reference bull genotypes36. In India, however, no such large dataset is available, owing to the large number of recognised cattle breeds (58), each having unique genotypic and phenotypic characteristics. Although smaller than the reference populations available in major international dairy breeding programs, the present dataset represents one of the most comprehensively recorded genomic resources currently available for Sahiwal cattle. Adjustment of effect of non-genetic factors is important for accurate and unbiased estimates of genetic parameters37. It was seen that the milk yield of Sahiwal cattle was greatly impacted by the time of birth. The effect was also reported significant by various researchers37–40. This might be due to the changing management of the farm with the time. Season of calving was found to have non-significant influence on the 305 days milk yield which was in line with the findings of some researchers32,38. Contrary to present study, significant effect of season of calving has been reported by Kumar et al.41. However, there might be no significant effect in present study as round the year similar management practices were followed including round the year fodder production in consultation with nutrition division thus resulting in similar nutritional management throughout year. The first lactation 305 days milk yield was significantly influenced by age at first calving which was in line with some researchers42,43 while others reported it as non-significant32,44. For the estimation of the genetic estimates we used the Single-Step Genomic Best Linear Unbiased Predictor (ssGBLUP) as it predicts an individual’s genomic estimated breeding value (GEBV) by combining genomic and pedigree data45–47. As ssGBLUP combines genomic and pedigree data, it is reported to increase the accuracy of genomic evaluation in many species as compared with pedigree-based BLUP or genomic BLUP48–51. Although this method has been effectively used in other nations for the genetic assessment of other species, no research of this kind on Sahiwal cattle has been conducted in India. The heritability estimate obtained in this study for 305 days milk yield (0.177 ± 0.068) was similar to those reported by several studies, whose estimates ranged from 0.14 to 0.2831,32,52,53. This heritability score indicates that this trait has additive genetic variance; as a result, utilizing it as a selection criterion can result in genetic benefits. It is important to interpret the results within the context of the nucleus breeding population used in the present study. As the herd has been subjected to long-term selection, the frequency of favourable alleles may be higher than that in the broader base population, potentially influencing estimates of genetic variance and heritability. However, the ranking of animals for milk yield is expected to remain largely unaffected. Furthermore, the incorporation of genomic information through ssGBLUP improved the accuracy of breeding value estimation and demonstrated the practical utility of genomic evaluation in a well-characterized Sahiwal breeding population. The findings of the present study provide an important foundation for the implementation of genomic evaluation and future genomic selection programs in this indigenous dairy breed. Further, the genomic estimated breeding values (GEBVs) were estimated for each individual and its accuracy was calculated using the methodology as described in the previous section. Prior estimates of breeding value has been made using ssGBLUP for a wide variety of statistical models in species such as buffalo54, dairy cattle55 and beef cattle48,50, laying hens56, broiler chickens57, turkey58, goat59, pigs60, sheep61,62, rainbow trout63, catfish64, honey bees65, horses66, and rabbits67 among others. We used single trait animal model of the ssGBLUP for the estimation of various genetic parameters in the Sahiwal cattle for which it is being used for the first time. The mean accuracy of the GEBVs estimated from prediction error variance was 43.11%, compared with 40.88% obtained using pedigree-based BLUP, indicating an improvement following the incorporation of genomic information. In addition, cross-validation analyses also demonstrated superior predictive performance of ssGBLUP (76.82%) relative to PBLUP (70.50%). The consistency of these results across both evaluation approaches supports the utility of genomic information for improving breeding value prediction in Sahiwal cattle. The accuracy of various genomic selection methods varies as per trait, population and method of estimation of GEBV. It varies 0.18 in Israeli Holstein68 to 0.54 in Chinese Holstein69 for the milk yield estimated by ssGBLUP. Further, it varies based on the method of estimation of GEBV like it gave value of 0.18 when using ssGBLUP for the milk yield while using Bayes C it gave value of 0.20 for the same trait68. Similar accuracy of breeding value for the milk production were achieved in the Saanen and Alpine goat breeds which were 49% and 45%, respectively70. However, Khan and co-workers have reported higher breeding value prediction accuracy for ssGBLUP (65%) in the crossbred cattle71 however this can be attributed to their large genomic dataset which consisted of 2572 animals. The effective population size (Ne), the number of reference population records (N), and the trait’s heritability are the primary factors that determine accuracy72. So, we further calculated the net effective population size of our Sahiwal herd. Numerous publications have acknowledged that linkage disequilibrium (LD) data could be used to estimate Ne73,74. We utilised SNeP1.11 which estimates historic effective population size (Ne) by using the relationship between linkage disequilibrium (r2), recombination rate (c), and Ne, which is based on the formula \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{E}}\left( {{{\mathrm{r}}^2}} \right)=\frac{1}{{1+4{N_e}c}}$$\end{document}75. Further, we also estimated Ne using the software program GONE24 and found out the Ne for the previous generation to be 71.927 which is above the current guidelines’ recommendation of Ne > 50 to avoid short-term inbreeding but is below recommended level of Ne > 500 to preserve long-term adaptive potential. It was in consistency with the Ne observed (Ne = 74) in Sahiwal cattle by other researchers76. Further, Ne of our Sahiwal herd is higher than the Sahiwal herd of the neighbouring country of India which is about 30 animals77. In our herd management the decision of selection and culling were taken by using the EPD of the animal. So, we further calculated EPD for maternal milk for each sire individually and was compared with the GEBVs calculated earlier using ssGBLUP methodology. It was seen that the animals with the higher GEBVs had also the higher EPD thus suggesting the GEBV can also be used as an aid for the selection and culling decisions to be undertaken at the farm. Furthermore, in our study we were able to select animals as early as 2 years of age (Sire 3021 M) but this interval can further be shortened by earlier collection of DNA sample from the animal and in practise it can be as early as the time of birth. Thus, genomic selection can potentially be used to increases the genetic gain per year by decreasing the generation interval. Similar findings have also been reported in other studies, stating that with the use of genomic markers in genetic evaluations, the generation interval for bulls’ dams and sires is greatly shortened to roughly 2.5 years, compared to 7 and 4 years, respectively, in traditional progeny testing procedures12. Genomic selection is also being implemented in US by the US Department of Agriculture (USDA) and they have reported that reductions in parent age for marketed bulls are the primary driver of the gain in the rate of genetic improvement because they have essentially halved the generation interval, thereby doubling the rate. Since the US began using genetic evaluation in 2010, the average net merit improvement rate for Holstein bulls has almost doubled78. Similar improvement rates can also be achieved in India, if the genomic selection programmes are properly implemented. Way forward/ future endeavours The benefit of genomics is greatest for traits with low heritability and ones that can be observed only late in life (such as longevity) because the increase in reliability provided by genomics is greatest for those traits12. A genomic assessment for cows usually has the advantage of being available soon after birth and is more reliable than what could be obtained from their own ancestral records. As the reference population size grew from 500 to 100,000 animals, the accuracy of genomic predictions increased from 0.42 to 0.98, demonstrating a non-linear rise in accuracy72. So, increasing the number of animals under the reference population will further increase the effectiveness of genomic selection. However, this is the typical case in developing countries, where not many animals are genotyped and thus adding more animals with recorded phenotypes and genotypes to the reference population especially in breeds with limited numbers is challenging79. A possible remedy is across-breed prediction; however, when prediction equations created in one breed are transferred to another, differences in LD, allele frequencies, and linkage phase frequently result in erroneous GEBVs80. Another interesting way to improve GEBV accuracy without incurring extra costs for genotyping or phenotyping is to combine multiple breeds into a multi-breed reference population. Sahana and co-workers used this strategy in the NDRI previously and concluded that benefit of using a multi-breed reference population for genomic prediction is larger for a numerically smaller breed, which gained the most from a multi-breed reference population and reported as much as 24.6% improvement in accuracy of genomic prediction in Kankrej breed in their study81. In addition to genetic examination, genomics can be utilised for tracking inbreeding, determining breed composition, pedigree finding, and mating programmes82. Breed-specific markers can be used to evaluate the breed composition of crossbred animals83, and in 2016 USDA developed a breed base representation value to display the proportion of DNA contributed to an animal by each of the five breeds that were undergoing genomic analysis84. India can also work on similar lines. Furthermore, various important gene markers can also be identified like for alpha-lactalbumin gene and these markers can aid in selecting elite animals for the future breeding programme85. Finding carriers of unwanted recessive traits is another advantage of genomics. In imputation, phasing the genotype into maternal and paternal haplotypes is a standard procedure. A certain haplotype may be a recessive lethal if it is never discovered to be homozygous, despite the fact that homozygotes are predicted based on haplotype frequencies86. A haplotype influencing fertility is identified when carrier sires are found to have lower fertility87. In terms of adding genotyped animals, India is also making progress. 10,000 human samples have just had their whole genomes sequenced as part of the Genome India project, producing an extensive inventory of India’s distinct genetic diversity. A similar pattern can be observed in animal breeding, where organisations like as NDRI, NBAGR, NDDB, NIAB, and others have led the way. Unified genomic chips, such as the Mahish Chip for buffaloes and the IndiGau Chip for native cattle, were developed specifically to begin the genomic selection of high genetic merit animals in the country. These indigenous chips will further increase the number of unique variants which can be incorporated in the study in order to select animals with high genetic merit and to accelerate the genetic improvement of cattle and buffaloes in the country with more precision88. Conclusions This study tried to realise the impact of the implementation of genomic selection of animals in Indian conditions on the Sahiwal cattle. Utilizing genomic data facilitates genetic analyses with sparse pedigree information and supports the initiation of breed improvement programs aimed at enhancing productivity. The heritability estimate for the milk yield was 0.17 ± 0.06, indicating further scope for selection to enhance milk production. Furthermore, GEBVs obtained from ssGBLUP methodology was also in line with the conventionally used method of EPD thus indicating that GEBVs can also be used as an aid for carrying out selection and culling decisions in a farm. Additionally, the ssGBLUP methodology allows to check for pedigree errors, where family relationships are incorrectly recorded. Thus, in a standard animal breeding program, given the limitations in resources and incomplete pedigree, we suggest employing the ssGBLUP model for regular genomic assessment and the identification of suitable candidates to effectively carry out a genomic selection program. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1