BIF recommends the use of EPD

From BIF Guidelines Wiki


The use of Best Linear Unbiased Prediction (BLUP) methodology to estimate Expected Progeny Differences (EPD) has been the norm in the North American Beef Industry for decades. Indeed, the global livestock genetics community utilizes the BLUP framework to drive genetic change. However, the US Beef Industry has continued to make components of EPD, including adjusted phenotypes and in some cases rankings based on genomic predictors, available. Publishing EPD components creates an unfavorable environment whereby EPD components are directly compared to EPD by some users, and selection decisions are incorrectly made using both the component and the EPD. A classic example of this is the all too common use of a combination of Calving Ease Direct EPD, Birth Weight EPD, adjusted birth weight of the animal, and perhaps the genomic predictor for calving ease, to make selection decisions when the most effective and correct way to select for calving ease is the exclusive use of the Calving Ease Direct EPD.

Selection decisions made based on adjusted phenotypes are suboptimal when an EPD for the trait exists; this is because environmental effects also create variation in the phenotypic value. An extension of adjusted phenotypes is contemporary group ratios. These values ignore valuable information which comes from relatives and from genomic information. As a consequence, both adjusted phenotypes and ratios are less accurate predictors of genetic merit than an EPD.

After nearly a decade of including genomic information into EPD, there still exists a desire by some to isolate the genomic-based prediction from the more traditional EPD, somehow rationalizing this decision believing one can learn more. The entire impetus of genomic prediction is to increase accuracy of genetic predictions and to avoid comparing two genetic predictors for the same trait. Publishing anything other than the complete EPD, that includes all available sources of information (pedigree, performance records, and genomic data), has the potential to decrease the accuracy of selection decisions and erode confidence in the process of genetic prediction.


BIF recommends that when an EPD is available, an adjusted phenotype and ratio should NOT be made available as it suggests validity in comparing EPD and adjusted phenotypes, and ratios and can lead to incorrect selection decisions. Further, BIF recommends that when genomic information is used to produce an EPD, a separate prediction based only the genomic information, or a ranking based on such a prediction, should NOT be made available as it decreases the accuracy of the decisions made when jointly using the genomic predictor and EPDs, and it creates confusion about the value and utility of genomic data.