Genomic predictions have revolutionized the dairy cattle industry by allowing for more accurate selection of breeding candidates and faster genetic progress. These predictions leverage genomic information, such as DNA markers, to estimate an animal’s genetic potential for various traits, such as milk production, fertility, and disease resistance. However, like any predictive model, genomic predictions for dairy cattle are susceptible to biases that can impact their reliability and effectiveness. In this article, we delve into the specific challenges and potential solutions related to bias in genomic predictions for dairy cattle.

Sources of Bias in Genomic Predictions for Dairy Cattle

1. Population Structure: Dairy cattle populations often exhibit complex genetic structures due to differences in breeds, lines, and geographical regions. When genomic prediction models are trained on data that does not adequately represent the target population, it can lead to bias. For instance, if a model is primarily developed using data from Holstein cattle and applied to Jersey cattle, it may produce biased predictions.
2. Marker Density and Quality: The quality and density of the genetic markers used in genomic predictions can introduce bias. Missing or erroneous marker data, as well as variations in genotyping platforms, can affect the accuracy of predictions.
3. Selective Breeding: In dairy cattle breeding, elite animals are often selected for reproduction based on their superior traits. If the training dataset primarily comprises these elite individuals, it may not accurately represent the broader population, leading to biased predictions.
4. Environmental Effects: Genomic prediction models typically assume that genetic effects are consistent across different environments. However, variations in management practices, nutrition, and climate can lead to genotype-environment interactions, resulting in biased predictions when cattle are transferred to different environments.
5. Sex-Biased Data: In some cases, data collection may be skewed towards one sex, such as focusing predominantly on cows while neglecting bulls. This sex bias can influence predictions for traits that exhibit sexual dimorphism.

Genetic trends of body or structural traits and dairy form for Holstein bulls. STA = standardized PTA; GPTA14 = genomic PTA in 2014; GPTA18 = genomic PTA in 2018; YOB = year of birth.

Mitigating Bias in Genomic Predictions for Dairy Cattle

Addressing bias in genomic predictions for dairy cattle is crucial for improving the accuracy of breeding decisions. Here are some strategies to mitigate bias:

1. Account for Population Structure: Employ statistical methods that account for population structure, such as principal component analysis (PCA) or genomic relationship matrices. These methods can correct for genetic stratification and reduce bias.
2. Marker Quality Control: Implement strict quality control measures to filter out low-quality markers and correct for genotyping errors. Consistency in marker data quality across datasets is essential.
3. Diverse Training Data: Aim to include a diverse set of animals in the training dataset, representing various breeds, lines, and geographical regions. This diversity can help improve prediction accuracy for a broader range of cattle.
4. Crossbred Animals: When possible, include crossbred animals in the training dataset to account for genetic differences between purebred populations. Crossbred animals can provide valuable information about heterosis and breed complementarity.
5. Environment-Specific Models: Develop environment-specific prediction models when relevant. Accounting for genotype-environment interactions can improve the accuracy of predictions when animals are moved to different management or climate conditions.

Genomic predictions have reshaped the dairy cattle industry by enhancing breeding decisions and genetic progress. However, the bias in these predictions remains a significant challenge. Dairy cattle breeders and researchers must be vigilant in addressing bias by considering population structure, marker quality, and environmental effects. By implementing these strategies, the dairy industry can harness the full potential of genomic predictions to enhance the genetic merit of their cattle, ultimately leading to improved productivity and sustainability.

This research estimated dairy cow feed efficiency genetic characteristics using random regression models. The study collected DMI, ECM, and MBW data from 7,440 first-lactation Holstein cows in six countries from January 2003 to February 2022. A multiple-trait random regression model using a fourth-order Legendre polynomial for all characteristics determined genetic parameters. Genomic residual feed intake (gRFI) was calculated from weekly DMI phenotypes using linear regressions on ECM and MBW to quantify feed efficiency genetically adjusted for these factors. Heritability estimates for DMI, ECM, MBW, and gRFI ranged from 0.15 (0.03) to 0.29 (0.02), 0.24 (0.01) to 0.29 (0.03), 0.55 (0.03) to 0.83 (0.05), and 0.12 (0.03) to 0.22 (0.06)

The findings improve knowledge of genetic parameter change during the first lactation and provide feed efficiency selection techniques for breeding programs. As feed prices rise, dairy cows may be able to improve feed efficiency during ideal lactation phases. The genetic link between feed efficiency indicator features must be understood to determine the best selection program attributes.

This research used random regression to estimate variance components and genetic factors across days in milk (DIM) of first-lactation Holstein cows for DMI, ECM, MBW, and genomic residual feed intake (gRFI). The research employed Resilient Dairy Genome Project data from pre-existing databases. The data on 7,440 first-lactation Holstein cows from six nations included 121,226 DMI, 120,500 ECM, and 98,975 MBW records. The amount of recordings from each source varied throughout a 305-day lactation. The pedigree file for phenotyped cows comprised up to 10 generations, totaling 30,776 animals. WOL daily measurements were averaged to generate weekly phenotypes. Due to biological limits, milk production less than 4kg, fat and protein yield larger than zero, MBW and DMI calculations, and at least 3 WOL records for a characteristic were removed during data editing. The data were adjusted to the Canadian mean and standard deviation owing to country heterogeneity.

The research estimated additive genetic, permanent environmental (PE), and residual variances for dairy milk yield (DMI), early calving months (ECM), and milk yield using a multiple-trait random regression animal model. Average Information Restricted Maximum Likelihood (AIREML) in WOMBAT computed variance components. All studies employed a fourth-order Legendre polynomial since it suited orders one through four best. The data analysis showed significance for all model parameters. All characteristics utilize the generic model:

y = Xb + Za + Wpe + e, where y is the vector of phenotypic records, X, Z, and W are the incidence matrices for fixed (b), additive genetic (a), and PE (pe) effects, and e is the vector of random residuals. Heritability estimates for several WOLs were determined using diagonal elements of G and P, as well as the homogeneous σ of each trait. If traits had correlated residuals, the residual covariance structure for multiple trait analysis was as follows.

The research estimated additive genetic, PE, and phenotypic WOL correlation standard errors using Robertson’s technique. The heritabilities, PE ratio, and phenotypic variance calculated the correlations. Genetic covariance components from multiple-trait random regression computed feed efficiency. ECM and MBW partial regression coefficients for each WOL were determined using the genetic covariance matrix (GCOVj). The equations were:

P.C.E.C.M=(G.C.O.V.12×G.C.O.V.23) – (G.C.O.V.13×G.C.O.V.22) – (12×G.C.O.V.12) – (11×G.C.O.V.22)
P, C, M, B, W = (G, C, O, V, 12 x 13) – (G, C, O, V, 11 x 23).
P(C:O:V:12 × G:C:O:V:12) − (G:C:O:V:11 × G:C:O:V:22) P(C:O:V:12 × G:C:O:V:13) −(G:C:O:V:11 × G:C:O:V:22) P(C:O:V:12 × G:C:O:V:12)

The research examined the heredity of dry matter intake, energy adjusted milk, metabolic body weight, and feed efficiency in first-lactation cows. The partial regression coefficients were utilized to linearly translate DMI phenotypes, which were corrected for energy sinks based on genetic connection. Adjusted DMI (gRFI) accounts for energy sinks’ genetic relationships.

Moderate heritability estimates for DMI, ECM, MBW, and gRFI. Similar DMI heritability estimates were published by Byskov et al., 2017, Li et al., 2018, and Krattenmacher et al., 2019. They evaluated ECM heritability within the range previously predicted. Heritability estimations for MBW were comparable to this research.

The heritability estimates for gRFI were 0.12 (0.02) to 0.23 (0.07). Previous research estimated gRFI heritabilities between 0.10 and 0.25. This research employed change in body weight, body weight, fat-protein adjusted milk, and milk energy, while prior investigations used somewhat different energy sinks.

Trait correlations were examined to find biological trends throughout lactation phases. It has been claimed that feed intake varies genetically during lactation. Changes in genetic and phenotypic correlations within a characteristic were examined through lactation to find these biological patterns.

DMI, ECM, MBW, and gRFI varied throughout early, mid, and late lactaction. Stronger additive genetic connections for DMI were seen when WOL was near together, with correlations ranging from 0.31 (0.17) to 0.99 (<0.01) for weeks far apart. However, correlations diminished with lactaction, notably significantly low DMI correlations between early and late lactation.

Like DMI, gRFI showed association with lactation phases from 0.04 (0.16) to 0.99 (0.01). Early and mid-lactation had the lowest associations, whereas later lactation had substantial relationships. ECM exhibited stronger lactation-wide correlations than DMI, but still showed lactation-specific variance. Timing of trait evaluation throughout lactation may affect selection to increase milk production and gRFI, therefore ECM alterations should be considered.

For all variables, WOL phenotypic correlations varied over lactation, with highest correlations between WOL closest together and lowest between WOL furthest away. Phenotypic associations were highest between WOL near together and lowest between WOL far apart. DMI has low to moderate associations during lactaction, suggesting that it should be regarded a characteristic at different periods. DMI’s dynamic behavior affects gRFI’s phenotype, hence this dynamic behavior may be extended to gRFI.

This research investigates genetic links between cows’ feed intake (DMI) and milk production (ECM) during lactation. DMI and ECM showed minor connections in early lactation but moderate to strong correlations by mid to late lactation. The change to higher favorable connections occurred when lactation intake met output needs. However, an increase in production without an increase in DMI, particularly in early lactaction, may prolong negative energy balance, which may harm health and fertility.

The genetic connection between metabolic body weight (MBW) and ECM changes from low positive to low negative during lactaction. This shows that bigger cows produce more milk early in lactation. The fast shift in correlations to weak negative correlations may indicate that livestock metabolism alters to gain weight as milk output drops and pregnancy progresses.

All characteristics had minimal phenotypic associations across lactation, with DMI and ECM in early lactation and MBW in late lactation. All WOLs had minimal phenotypic correlations between gRFI and MBW and ECM. However, DMI and gRFI displayed high positive correlations in comparable WOLs, indicating DMI’s dynamic phenotypic behavior and effect on gRFI.

In conclusion, breeding for gRFI requires knowing the link between DMI, MBW, ECM, and characteristics like energy balance and body condition score.

The Scandinavian indexes (Denmark, Sweden, and Finland) are now accessible ONLINE. Mecanico from the United States has the highest NTM genetic sire this August 2023 proof, with +47 NTM. VH Karat *RC comes in second at +44 NTM, followed by VH Sweet P at +41 NTM. VH Deco *RC, together with the German sire Youngester, again tops the daughter-proven NTM list with +31 NTM. At +29 NTM, they are followed by three sires: Semino, VH Bahrain, and VH Goal.

• SCANDINAVIA – NTM Genomic bulls 08/23
• SCANDINAVIA – NTM Proven bulls 08/23

This August 2023 run, Genosource Captain came in as the #1 gPLI daughter proven sire with +807 gPLI and +1.65 TM (Type). He is followed by the Progenesis Wimbledon at +756 PLI, and Kimmel is third at +733 gPLI. Denovo Leeds leads the gPLI genomics sire index with +914 gPLI. He is followed by Denovo Highview at +902 PLI and Wilra SSI Faneca Ebersol at +896 PLI, rounding out the top three.

• UK – PLI Young Genomic Bulls – 08/23
• UK – PLI Daughter Proven Bulls – 08/23
• UK – TYPE Merit Young Genomic Bulls – 08/23
• UK – TYPE Merit Daughter Proven – 08/23 

The new Italian indices have just arrived. Zingler son Go-Farm Abacuc leads the genomic PFT Domestic ranking with +5146 gPFT. More than 100 points down, we discover another Zingler son, Isolabella Inseme Prinston at +5008 PFT, rounding out the top three, which is completed by Zingler son Bardo at +4983 gPFT. Topping in the daughter-proven domestic PFT list is a Rio son with +4417 gPFT, he is followed by Perseus son Gaudi at +4400 gPFT and at the third slot we have Go-Farm Bernini at +4344 gPFT.

• ITALY – Dtr proven PFT (Domestic) – 08/23
• ITALY – Genomic PFT (Domestic) – 08/23
• ITALY – Foreign Genomic PFT – 08/23

Progenesis Partner (s. Mahomes) at +1646 ISET is the new leader on the Swiss chart. He is followed by Swissgen Lewitan at +1642 ISET, who rounds off the stage with 3star HWH Globe at +1631 ISET. Wilra SSI Rivet Genuine leads the Interbull daughter proven list with +1581 ISET, followed by S-S-I Hodedoe Montley with +1561 ISET and Larcrest Commitment with +1556 ISET. NH DG Arrow at AI-Total continues to be the  #1 daughter confirmed ITP (TYPE) sire with >113 Gliedmassen (F&L) and >0.07% PROTEIN. He is the #14 daughter-proven ITP (Type) sire in the world and the #5 daughter-proven ITP (Type) sire with more than 1280 ISET.

• SWITZERLAND – ISET – Genomic 08/23
• SWITZERLAND – ISET – Interbull Dtr proof – 08/23
• SWITZERLAND – ISET – Domestic dtr proven 08/23
• SWITZERLAND – Top 100 ISET heifers 08/23
• SWITZERLAND – Top 100 ISET cows 08/23
• SWITZERLAND – Top 500 RC RED ISET heifers 08/23