Estimation & Prediction

National Across-Region Evaluation of Milk Yield in Indian Dairy Cattle and Buffaloes

---

Abstract

India's dairy production spans highly diverse management systems and environments, complicating uniform national genetic evaluations. Progeny testing projects (hereafter referred to as "projects") for Murrah and Mehsana buffaloes and crossbred Holstein Friesian x Indicus (CBHF) and Jersey x Indicus (CBJY) cattle have been implemented in multiple regions since the 1980s. Each such project is implemented by an agency involved in bovine productivity enhancement such as livestock development board or milk federation or trust-based institutions constituted by an act of government. Bos indicus breeds in CBHF can be either of Sahiwal/Gir/Kankrej while those in CBJY can be either of Sahiwal/Red Sindhi/Gir. Joint evaluations across 2-4 such projects per breed caused bias, with bulls from projects in favorable environments ranking higher. Also, due to biased variance estimates when including non-pedigreed animal data, analyses were restricted to first-lactation daughters only. This study developed a National Across-Region Evaluation (NARE) framework to jointly analyze data of the same breed, improving the accuracy and country-wide relevance of genomic estimated breeding values (GEBVs) while indirectly accounting for genotype-by-environment interactions. Two key modifications were implemented within a single-breed ssGBLUP random regression model: (i) Nested days-in-milk regression by project/geography, allowing each project an independent lactation curve (Legendre polynomials) and (ii) Harmonized project-specific fixed effects, preserving within-project management differences while enabling a unified analysis. To incorporate non-pedigreed yet phenotyped animals without inflating variance estimates, a two-step approach was adopted. In step 1, fixed effects were estimated using the full dataset and used to derive corrected yields. In step 2, only pedigreed/genotyped individuals were used for variance component and GEBV estimation, with corrected yield as the response. Model performance was assessed through 10-fold cross-validation, comparing NARE with individual project analyses. Predictive ability and potential bias were evaluated by regression of corrected yield on GEBVs and by rank correlations of bulls across projects. Joint evaluations increased correlations between GEBVs and corrected yield by around 35-40%, particularly in projects with smaller datasets or shallower pedigrees. Rank correlations between NARE and individual project analyses broadly ranged from 0.60 to 0.90, indicating moderate re-ranking but improved stability. Importantly, top bulls were more evenly represented across projects, demonstrating reduced environmental bias and enhanced comparability. The NARE framework effectively integrates data from geographically diverse PT projects through model adjustments and a two-step estimation strategy. It enables inclusion of non-pedigreed phenotypes, improves predictive accuracy, and supports joint sire ranking across regions without major loss of within-project specificity. This approach provides a scalable template for national-level genomic evaluations in heterogeneous production systems such as India's.

Keywords: 2026

How to Cite:

Gajjar, S., Nayee, N., Sudhakar, A., Kesharwani, A., Mahajan, A., Gupta, R. & Trivedi, K., (2026) “National Across-Region Evaluation of Milk Yield in Indian Dairy Cattle and Buffaloes”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286871. doi: https://doi.org/10.31274/wcgalp.24149

Rights: 1