J. Dairy Sci. 104:5111–5124
`https://doi.org/10.3168/jds.2020-19777
`© 2021, The Authors. Published by Elsevier Inc. and Fass Inc. on behalf of the American Dairy Science Association®.
`This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
`Invited review: The future of selection decisions and breeding
`programs: What are we breeding for, and who decides?
`John B. Cole,1*† João W. Dürr,2 and Ezequiel L. Nicolazzi2
`1Animal Genomics and Improvement Laboratory, Agricultural Research Service, United States Department of Agriculture (USDA), Beltsville,
`MD 20705-2350
`2Council on Dairy Cattle Breeding, 4201 Northview Drive, Suite 302, Bowie, MD 20716
`
`ABSTRACT
`
`Genetic selection has been a very successful tool
`for the long-term improvement of livestock popula-
`tions, and the rapid adoption of genomic selection over
`the last decade has doubled the rate of gain in some
`populations. Breeding programs seek to identify geneti-
`cally superior parents of the next generation, typically
`as a function of an index that combines information
`about many economically important traits into a single
`number. In the United States, the data that drive
`this system are collected through the national dairy
`herd improvement program that began more than a
`century ago. The resulting information about animal
`performance, pedigree, and genotype is used to com-
`pute genomic evaluations for comparing and ranking
`animals for selection. However, the full expression
`of genetic potential requires that animals are placed
`in environments that can support such performance.
`The Agricultural Research Service of the US Depart-
`ment of Agriculture and the Council on Dairy Cattle
`Breeding collaborate to deliver state-of-the-art genomic
`evaluations to the dairy industry. Today, most breeding
`stock are selected and marketed using the net merit
`dollars (NM$) selection index, which evolved from 2
`traits in 1926 (milk and fat yield) to a combination of
`36 individual traits following the last NM$ update in
`2018. Updates to NM$ require the estimation of many
`different values, and it can be difficult to achieve con-
`sensus from stakeholders on what should be added to,
`or removed from, the index at each review, and how
`those traits should be weighted. Over time, the major-
`ity of the emphasis in the index has shifted from yield
`traits to fertility, health, and fitness traits. Phenotypes
`for some of these new traits are difficult or expensive
`
`Received October 13, 2020.
`Accepted January 3, 2021.
`*Current address: URUS Group LP, 2418 Crossroads Drive, Suite
`3600, Madison, WI 53718.
`†Corresponding author: john.b.cole@ gmail .com
`
`to measure, or require changes to on-farm habits that
`have not been widely adopted. This is driving interest
`in sensor-based systems that provide continuous mea-
`surements of the farm environment, individual animal
`performance, and detailed milk composition. There is
`also a need to capture more detailed data about the
`environment in which animals perform, including in-
`formation about feeding, housing, milking systems, and
`infectious and parasitic load. However, many challenges
`accompany these new technologies, including a lack
`of standardization or validation, need for high-speed
`internet connections, increased computational require-
`ments, and interpretations that are often not backed by
`direct observations of biological phenomena. This work
`will describe how US selection objectives are developed,
`as well as discuss opportunities and challenges associ-
`ated with new technologies for measuring and recording
`animal performance.
`Key words: breeding programs, genetic improvement,
`selection objectives, total merit indices
`
`
`
`INTRODUCTION
`
`Selection indices are essential tools in modern dairy
`cattle breeding because they enable information about
`many traits to be combined into a single value for rank-
`ing animals and making selection decisions. The ideal
`breeding objective for dairy cattle remains a popular
`topic, even if consensus is elusive, and is frequently
`discussed in the scientific and popular literature (e.g.,
`Hazel et al., 1994; Philipsson et al., 1994; VanRaden,
`2004; Miglior et al., 2005; Shook, 2006; Miglior et
`al., 2017; Cole and VanRaden, 2018; Binversie, 2019;
`Dechow, 2020; Schmidt, 2020). There is no single selec-
`tion objective that is ideal for all populations, or all
`herds within a population, but there is a general set of
`principles that should be followed when developing an
`index (e.g., Cameron, 1997).
`Historically, selection indices in the United States
`were developed by the United States Department of
`Agriculture (USDA) and by purebred dairy cattle
`
`5111
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5112
`
`associations (PDCA). Input has also been provided
`by scientists at land-grant universities and technical
`experts at breeding companies, using data available
`through the national milk recording system and breed
`type appraisal programs. Proposed indices from the
`USDA were typically reviewed by groups of experts,
`and information about the derivation of the indices was
`published in technical and trade publications, ensuring
`confidence in the values because of that review process.
`Recently, genetic evaluations for novel traits and new
`selection indices have been computed and distributed
`by companies such as GENEX and Zoetis. Both of these
`organizations publish their own indices, which include a
`combination of traits from the Council on Dairy Cattle
`Breeding (CDCB) evaluations and their own propri-
`etary traits (i.e., hoof health in the case of GENEX,
`and cow and calf health traits in the case of Zoetis).
`This provides farmers with new tools and may drive
`demand for new phenotypes, but transparent review
`processes are often lacking. Correlations among indices
`are generally strong (T. J. Lawlor Jr., Holstein Associa-
`tion USA, Brattleboro, VT; personal communication),
`and in such cases, it is unclear if new tools provide new
`information or serve only as marketing tools.
`This paper will describe how decisions about selec-
`tion indices are made in the United States, discuss
`traits that may be included in future changes to exist-
`ing indices, and identify opportunities associated with
`new technologies for recording animal performance.
`Although the focus is on the US dairy sector, examples
`from other countries are discussed when appropriate.
`
`DEVELOPMENT OF SELECTION OBJECTIVES
`Who Are the Participants in the US Dairy Sector?
`
`To explain how selection decisions are made, we must
`briefly review the stakeholders in the process (Wiggans
`et al., 2017; Figure 1). The Animal Genomics and Im-
`provement Laboratory is part of the Agricultural Re-
`search Service, USDA’s in-house research arm, and was
`responsible for the development of the indices shown in
`Table 1 (sometimes under other laboratory names due
`to Agricultural Research Service organizational chang-
`es). The CDCB operates the national genetic evalua-
`tion system and maintains the national cooperator da-
`tabase. The CDCB board includes representatives from
`all key industry participants, including the National
`Dairy Herd Information Association (NDHIA), Dairy
`Records Processing Centers, the National Association
`of Animal Breeders, and the PDCA. The field service
`organizations and milk testing laboratories that operate
`the national milk recording program are represented by
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`NDHIA; the organizations that aggregate and distrib-
`ute milk testing data and provide herd management
`information are represented by the Dairy Records Pro-
`cessing Centers; the AI companies, who own most of
`the bulls and many elite females, are represented by
`the National Association of Animal Breeders; and the
`breeders, who own most of the elite cattle, are repre-
`sented by the PDCA. In addition, CDCB has several
`advisory groups that include farmers, researchers, and
`allied industry personnel that review and provide feed-
`back on data quality and proposed changes to the ge-
`netic evaluation system. Scientists from the land-grant
`universities provide valuable technical expertise to the
`Animal Genomics and Improvement Laboratory and
`CDCB, both as individual consultants and through the
`SCC-084 Multistate Research Coordinating Committee
`and Information Exchange Group. This group meets
`annually to share results and plan future research on
`selection and mating strategies to improve dairy cattle
`performance, efficiency, and longevity. All of these par-
`ticipants in the national dairy improvement program
`have opportunities to influence the selection indices
`adopted by CDCB, some directly, and others indirectly.
`
`Figure 1. The general structure of the US dairy cattle improve-
`ment sector. Solid lines indicate board membership in an organization,
`and broken (dashed) lines represent advisory relationships. AGIL =
`Animal Genomics and Improvement Laboratory, Agricultural Research
`Service, United States Department of Agriculture (constructs the in-
`dex); CDCB = Council on Dairy Cattle Breeding (operates the nation-
`al genetic evaluation system and maintains the national cooperator
`database); DRPC = Dairy Records Processing Centers (aggregate and
`distribute milk testing data and provide herd management informa-
`tion); NAAB = National Association of Animal Breeders (represents
`breeding companies); DHI = Dairy Herd Improvement (oversees the
`national milk recording program); PDCA = Purebred Dairy Cattle
`Associations (represents breeders). Scientists at the land-grant univer-
`sities provide technical expertise.
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5113
`
`In addition to the organizations with direct represen-
`tation on the CDCB board, there are several entities
`that participate in the collection and transfer of genom-
`ic information (Figure 2). The genomic nominators are
`responsible for collecting DNA samples from the animal
`owner, providing CDCB with information about the
`animals sampled, and transferring the DNA samples to
`the genotyping laboratory. The genotyping laboratory
`extracts DNA from samples, prepares SNP genotypes,
`provides summary information back to the nominator,
`and transfers the genotypes to the CDCB. Genomic
`evaluations are sent from the CDCB to the nominators,
`and on to the records providers. Both the nominators
`and laboratories must meet quality certification guide-
`lines before they are permitted to participate in the
`system, and their performance is audited on an annual
`basis.
`
`How Are Decisions About Selection Criteria Made?
`
`How Are the Index Weights Determined? Selec-
`tion indices must be periodically updated to include
`new traits and reflect changing economic conditions,
`as well as changing genetic parameters between and
`among traits. From the development of the first USDA
`index (Norman and Dickinson, 1971) until the pres-
`ent (VanRaden et al., 2018), a collaborative model has
`been used to propose and adopt changes to the indi-
`ces. Although an argument can be made that changes
`
`should be driven strictly by mathematics—and we are
`sympathetic to this position—the reality is that tools
`will not be adopted unless the intended users perceive
`value in the tool. The net merit dollars (NM$) weights
`are primarily based on selection index theory, with
`fine-tuning based on consensus expert opinion, which
`reflects the well-known challenge of computing index
`weights (Freeman, 1984). It is also more difficult to
`compute the incomes and expenses associated with
`traits in the index than the textbooks suggest, and
`input from the field is very helpful in that regard. Our
`experience over the last 50 yr suggests that collabora-
`tion not only drives increased adoption of the indices,
`it also builds support for other communal efforts, such
`as the recording of new phenotypes so that they may
`eventually be included in the index.
`Who Owns the Index? Responsibility for the na-
`tional cooperators database and the genetic evaluation
`system was passed from USDA to CDCB in 2013, but
`NM$ and its companion indices (cheese merit, fluid
`merit, and grazing merit; VanRaden, 2000; Gay et al.,
`2014) require both index weights and genetic values
`to compute. When an index is owned by a PDCA or
`an AI company, it is clear who has the authority to
`make changes and the responsibility for distributing
`the calculations. In the case of NM$, USDA and CDCB
`share these roles: USDA is responsible for construction
`of the index, and CDCB provides the data needed to
`calculate and distribute the values. Both organizations
`
`Table 1. Traits included in United States Department of Agriculture selection indices1 and the relative emphasis placed on each, 1971–2018
`(Cole and VanRaden, 2018)
`
`Relative emphasis on trait (%)
`
`Trait2
`
`PD$
`(1971)
`
`MFP$
`(1976)
`
`CY$
`(1984)
`
`NM$
`(1994)
`
`NM$
`(2000)
`
`NM$
`(2003)
`
`NM$
`(2006)
`
`NM$
`(2010)
`
`NM$
`(2014)
`
`NM$
`(2017)
`
`NM$
`(2018)
`
`27
`46
`27
`
`–2
`45
`53
`
`−1
`−1
`−1
`0
`0
`0
`5
`6
`52
`Milk
`27
`24
`22
`19
`23
`22
`21
`25
`48
`Fat
`17
`18
`20
`16
`23
`33
`36
`43
`—
`Protein
`12
`13
`19
`22
`17
`11
`14
`20
`—
`—
`—
`PL
`−4
`−7
`−7
`−10
`−9
`−9
`−9
`−6
`—
`—
`—
`SCS
`7
`7
`8
`7
`6
`7
`7
`—
`—
`—
`—
`UC
`3
`3
`3
`4
`3
`4
`4
`—
`—
`—
`—
`FLC
`−5
`−6
`−5
`−6
`−4
`−3
`−4
`—
`—
`—
`—
`BWC
`7
`7
`7
`11
`9
`7
`—
`—
`—
`—
`—
`DPR
`—
`—
`—
`—
`—
`−2
`—
`—
`—
`—
`—
`SCE
`—
`—
`—
`—
`—
`−2
`—
`—
`—
`—
`—
`DCE
`5
`5
`5
`5
`6
`—
`—
`—
`—
`—
`—
`CA$
`1
`1
`1
`—
`—
`—
`—
`—
`—
`—
`—
`HCR
`2
`2
`2
`—
`—
`—
`—
`—
`—
`—
`—
`CCR
`7
`7
`—
`—
`—
`—
`—
`—
`—
`—
`—
`LIV
`2
`—
`—
`—
`—
`—
`—
`—
`—
`—
`—
`HTH$
`1PD$ = Predicted Difference Dollars (Dickinson et al., 1971); MFP$ = Milk-Fat-Protein Dollars (Norman et al., 2010); CY$ = Cheese Yield
`Dollars (Norman, 1986); NM$ = Lifetime Net Merit Dollars (VanRaden and Wiggans, 1995).
`2PL = productive life; UC = udder composite; FLC = feet and legs composite; BWC = body weight composite; DPR = daughter pregnancy
`rate; SCE = sire (direct) calving ease; DCE = daughter (maternal) calving ease; CA$ = calving ability dollars; HCR = heifer conception rate;
`CCR = cow conception rate; LIV = cow livability; HTH$ = health dollars.
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5114
`
`Figure 2. Flow of information among participants in the national genomic evaluation system.
`
`have input into the evolution of the index, but neither
`owns it. Lifetime net merit was initially developed by
`USDA scientists (VanRaden and Wiggans, 1995), but
`it has remained relevant because of the USDA-CDCB
`partnership. The success of NM$ has not prevented
`other organizations from developing their own selection
`tools, and farmers have many indices from which to
`choose if NM$ does not meet their expectations.
`How Do We Validate Our Indices? Selection
`indices are constructed using many calculations based
`on a substantial body of scientific theory (e.g., Hazel
`et al., 1994). Complex traits, such as longevity, remain
`difficult to model properly, and there is some dispar-
`ity between management practices in the field and
`optimal economic strategies (Schmitt et al., 2019; De
`Vries, 2020). It can be difficult to confirm that realized
`selection gains are consistent with index predictions,
`but some recent reports show that animals with greater
`genetic merit are more profitable than their contem-
`poraries with lower rankings. In the United States,
`scientists from Zoetis and the University of Pennsyl-
`vania recently showed that cows in the top quartile for
`the Dairy Wellness Profit index had greater lifetime
`profit than herdmates in lower quartiles (Fessenden et
`al., 2020). The lack of farm-level income and expense
`data in the national cooperator database makes it dif-
`
`ficult to perform routine validation of the index, but
`collaboration with projects such as the Dairy Profit
`Monitor program at Cornell University (https: / / cals
`.cornell .edu/ pro -dairy/ our -expertise/ business/ dairy
`-profit -monitor) could support such an effort.
`
`How Have Selection Indices Evolved Over Time?
`
`What Traits Are in the Index? The emphasis
`placed on each trait in each revision of the selection
`index is shown in Table 1, and the rate at which new
`traits are added to the index has increased consider-
`ably in recent years. This represents changes in dairy
`economics, an improved understanding of the biology of
`the cow, and greater ease of collecting and transferring
`data. The first selection index published by USDA was
`the Predicted Difference Dollars index, which included
`information about milk and fat production (Norman
`and Dickinson, 1971). Although it was recognized at
`the time that other traits might have economic impor-
`tance, milk and fat were the only traits with enough
`phenotypic information available to support genetic
`evaluations. Protein yield was added to Predicted Dif-
`ference Dollars in 1976 to produce the Milk-Fat-Protein
`Dollars index (Norman et al., 1979), and an index for
`cheese yield was developed in 1984 (Norman, 1986).
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5115
`
`This was the status quo until 1994, when productive
`life and SCS were combined with the yield traits to
`produce the first iteration of the Lifetime Net Merit
`index (VanRaden and Wiggans, 1995).
`Although the combination of fitness, conformation,
`and production traits included in the first version of
`NM$ in 1994 set it apart from most of its international
`competitors, the Scandinavian countries began record-
`ing health and fertility data in the 1960s and computing
`genetic evaluations for those traits in the 1970s (Phil-
`ipsson and Lindhé, 2003). Their experience showed that
`selection objectives that include traits with low heri-
`tabilities can produce worthwhile gains in cow health
`and fertility. Leitch (1994) reviewed 19 contemporary
`selection indices and found that only 2 (Danish S-Index
`and US NM$) included mastitis resistance, 1 included
`fertility (Danish S-Index), and 1 included productive
`life (US NM$). In a review based on an independent
`survey, Philipsson et al. (1994) identified several other
`countries’ indices (Finland, Norway, Slovenia, and Swe-
`den) that also included fitness traits. When Miglior et
`al. (2005) revisited the subject a decade later, each of
`the 17 indices reviewed included 1 or more fitness traits
`as part of the selection criterion. This trend toward the
`inclusion of more fitness traits in total merit indices has
`continued (Cole and VanRaden, 2018), and it is now
`more remarkable when an index does not include such
`traits than when it does.
`There Is No Universal Standard. It is tempt-
`ing to assume that it is possible to define a universal
`total merit index, but that is not possible because every
`farmer operates in a slightly different economic and
`environmental setting than their neighbors. In theory,
`every farm should actually use its own selection index
`that is customized to its financial situation and busi-
`ness objectives (Gjedrem, 1972). In practice, farms
`with similar operating and financial characteristics can
`use the same index with little loss of efficiency. It is also
`difficult to assign direct economic values to some traits,
`most notably conformation traits. Different breeders
`have different goals, which can affect their breeding
`programs. A commercial dairy that derives its income
`principally from the sale of milk solids will have differ-
`ent incomes and expenses than a seedstock breeder who
`also sells embryos and elite germplasm, and they may
`benefit from using different indices. Lifetime net merit
`is explicitly developed for use by commercial dairy
`farmers (VanRaden, 2004), and Holstein Association
`USA’s Total Performance Index is intended for use by
`registered cattle breeders who often sell genetics as well
`as milk.
`More than 1 index is needed because farmers sell
`their products into different markets (e.g., VanRaden,
`
`2000), have different personal preferences (e.g., Martin-
`Collado et al., 2015), and strategies for maximizing
`profit vary (e.g., Berry et al., 2019). As noted earlier,
`the CDCB publishes 4 separate indices (lifetime net
`merit, fluid merit, cheese merit, and grazing merit)
`to provide farmers with options that best match their
`needs. The strategy of providing multiple indices to its
`farmers is certainly not unique to the United States For
`example, when the Australian Dairy Herd Improvement
`Scheme (now DataGene) revised the Australian Profit
`Ranking index in 2016, they replaced it with 3 new
`indices (Byrne et al., 2016). The Balanced Performance
`Index, Health Weighted Index, and Type Weighted
`Index allow their farmers to focus on trait groups that
`are most important to them within a technically sound
`framework.
`Are There Too Many Indices Already? The
`last several years have seen the development of many
`new selection indices marketed to commercial dairy
`farmers. In contrast to NM$ and indices published by
`PDCA, many of these new indices are promoted by
`breeding companies as a means of differentiating their
`products. Several selection indices currently available
`to US dairy farmers are shown in Table 2, although
`this is not an exhaustive list (some organizations do
`not make the details of their index publicly available).
`These tools include indices developed by USDA, PDCA
`(e.g., American Jersey Cattle Association), and com-
`mercial organizations (e.g., Zoetis). In general, most
`indices are similar in that they are seeking to find a bal-
`ance between productivity (the direct source of much
`farm income) and fitness traits (often a source of direct
`costs). Direct comparisons are challenging because
`some indices are available only for bulls marketed by
`the publisher of the index. Most differences among in-
`dices are due to the inclusion of different sets of traits,
`or to the differential weighting of such traits in the
`index. Some companies develop proprietary evaluations
`to differentiate their offerings from those of their com-
`petitors. Correlations among these indices generally are
`very strong, and there is minimal reranking of bulls
`when moving from one index to another (T. J. Lawlor
`Jr., Holstein Association USA, Brattleboro, VT; per-
`sonal communication). However, farmers may not be
`able to clearly describe differences between each index,
`providing some opportunities for confusion. There also
`is concern that marketers may be over-stating the im-
`portance of the differences between the indices.
`Are Selection Indices Responsible for Reduc-
`ing Diversity in Some Breeds? It is tempting to
`place the blame for the ongoing loss of genetic diver-
`sity in US Holsteins (e.g., Maltecca et al., 2020) on
`breeders who avidly pursue high-index animals, but
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5116
`
`Table 2. Some selection indices1 currently offered to US dairy farmers
`
`Trait2
`
`BS PPR
`(2017)
`
`AY CPI
`(2019)
`
`GU PTI
`(2020)
`
`JE JPI
`(2020)
`
`HO ICC$
`(2020)
`
`JE ICC$
`(2020)
`
`HO TPI
`(2020)
`
`USDA NM$
`(2018)
`
`−1
`27
`17
`12
`−4
`7
`3
`−5
`7
`—
`—
`
`
`5
`1
`2
`7
`2
`
`—
`—
`5
`—
`—
`—
`—
`Milk
`19
`22
`14
`19
`25
`25
`28
`Fat
`19
`22
`12
`27
`25
`35
`34
`Protein
`5
`12
`6
`5
`6
`—
`6
`PL
`−4
`−4
`−4
`−4.5
`—
`−4
`—
`SCS
`11
`5
`7
`—
`10
`—
`10
`UC
`6
`—
`—
`—
`10
`—
`—
`FLC
`—
`—
`—
`—
`—
`—
`—
`BWC
`9.1
`12
`8
`9
`15
`6
`12
`DPR
`—
`—
`−2
`—
`—
`—
`—
`SCE
`−0.5
`—
`−1
`—
`—
`—
`—
`DCE
`—
`—
`−1
`—
`—
`—
`—
`SSB
`−1.5
`—
`−1
`—
`—
`—
`—
`DSB
`—
`—
`—
`—
`—
`—
`—
`CA$
`1.3
`6
`5
`2
`—
`—
`—
`HCR
`1.3
`3
`—
`3.5
`—
`—
`—
`CCR
`3
`5
`2
`3
`3
`—
`4
`LIV
`2
`—
`4
`4.6
`—
`—
`—
`HLTH
`—
`—
`—
`—
`—
`—
`6
`MO
`
`
`8
`—
`—
`19.4
`—
`25
`—
`TYPE
`
`—
`—
`—
`—
`—
`5
`—
`UDEP
`
`—
`—
`—
`—
`3
`—
`—
`STR
`
`—
`—
`—
`—
`3
`—
`—
`STAT
`
`—
`—
`—
`−3
`—
`—
`—
`DENS
`
`8
`—
`16
`—
`—
`—
`—
`FEED
`
`—
`—
`1
`—
`—
`—
`—
`POLL
`
`—
`1
`<1
`—
`—
`—
`—
`HAPL
`
`—
`—
`6
`—
`—
`—
`—
`LOCO
`
`—
`—
`1
`—
`—
`—
`—
`HOOF
`
`—
`—
`1
`—
`—
`—
`—
`BCS
`
`—
`4
`1
`—
`—
`—
`—
`MAST
`
`—
`—
`1
`—
`—
`—
`—
`SPD
`
`—
`—
`<1
`—
`—
`—
`—
`TEMP
`
`—
`4
`—
`—
`—
`—
`—
`CALF
`
`1.3
`1
`—
`—
`—
`—
`—
`EFC
`1Due to rounding, columns will sometimes sum to a value slightly smaller or larger than 100. BS PPR = Brown Swiss Progressive Performance
`Ranking (Brown Swiss Association, 2017); AY CPI = Cow Performance Index (US Ayrshire Breeders’ Association, 2020); GU PTI = Performance
`and Type Index (American Guernsey Association, 2020); JE JPI = Jersey Performance Index (Tauchen, 2020); HO ICC$ = Ideal Commercial
`Cows for Holsteins (Genex, 2020a,b); JE ICC$ = Ideal Commercial Cows for Jerseys (Genex, 2020a,b); HO TPI = Total Performance Index
`(Holstein Association USA, 2020); USDA NM$ = Net Merit Dollars (VanRaden et al., 2018).
`2PL = productive life; UC = udder composite (varies by breed and index); FLC = feet and legs composite; BWC = body weight composite; DPR
`= daughter pregnancy rate; SCE = sire (direct) calving ease; DCE = daughter (maternal) calving ease; CA$ = calving ability dollars; HCR =
`heifer conception rate; CCR = cow conception rate; LIV = cow livability; HLTH = health traits (varies by breed and index); MO = mobility
`(Brown Swiss); TYPE = type (conformation) composite (varies by breed); UDEP = udder depth; STR = strength; STAT = stature; DENS =
`milk density; FEED = feed intake/feed cost (varies by breed and index); SSB = sire (direct) stillbirth; DSB = daughter (maternal) stillbirth;
`POLL = polled status; HAPL = haplotypes affecting fertility; LOCO = locomotion; HOOF = hoof health; MAST = clinical mastitis; SPD =
`milking speed; TEMP = milking temperament; CALF = calf survivability; EFC = early first calving (age at first calving).
`
`increases in rates of inbreeding are more likely driven
`by gains in selection intensity as a result of genomic
`technology (García-Ruiz et al., 2016). The short gen-
`eration intervals and high rates of gain in each genera-
`tion have driven seedstock producers to sample heavily
`within lines that have already produced successful
`bull families. Each company has limited resources for
`identifying elite animals, and the risk of losing market
`share to a competitor is greater now than it was un-
`der traditional progeny testing programs because of
`the speed with which genetic gains accumulate. This
`is probably why the decline in the rate of inbreeding
`
`under genomic selection predicted by Daetwyler et al.
`(2007) has not materialized—no large AI company is
`willing to risk sampling largely from outcross families.
`Even if there was a market for outcross bulls, most
`phenotypes are collected from daughters of popular
`families, and prediction accuracies will be lower for
`the outcross animals. However, the long-term value of
`broadening the genetic base may be worth the sacri-
`fice of some short-term accuracy. This is similar to
`the trade-off between changes in inbreeding and rates
`of genetic gain made in optimal contribution theory
`(Clark et al., 2013).
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5117
`
`Table 3. Hypothetical composition of a future lifetime Net Merit index that is the sum of subindices for production (PRO$), longevity (LON$),
`fertility (FER$), type (conformation; TYP$), calving ability (CA$), and health (HTH$)1
`
`PRO$ (39)
`
`
`
`LON$ (27)
`
`
`
`FER$ (10)
`
`
`
`TYP$ (11)
`
`
`
`CA$ (7)
`
`
`
`HTH$ (6)
`
`Milk (1)
`Fat (21)
`Protein (17)
`
` MAST (1)
`SCE (1)
`
`FLC (2)
`
`DPR (6)
`
`PL (12)
`
` METR (1)
`SCE (1)
`
`UDC (5)
`
`HCR (1)
`
`LIV (7)
`
`
`RPL (1)
`SSB (1)
`
`BWC (4)
`
`CCR (2)
`
`HLV (2)
`
` KETO (1)
`DSB (1)
`
`
`
`EFC (1)
`
`SCS (6)
`
`
`DSAB (1)
` GL (2)
`
`
`
`
`
`
` MFEV (1)
`
`BWT (1)
`
`
`
`
`
`
`1PL = productive life; BWC = body weight composite; UDC = udder composite; FLC = feet and legs composite; DPR = daughter pregnancy
`rate; SCE = sire (direct) calving ease; DCE = daughter (maternal) calving ease; SSB = sire (direct) stillbirth; DSB = daughter (maternal)
`stillbirth; HCR = heifer conception rate; CCR = cow conception rate; LIV = cow livability; GL = gestation length; RFI = residual feed intake;
`MFV = milk fever (hypocalcemia); DAB = displaced abomasum; KET = ketosis; MAS = clinical mastitis, MET = metritis; RPL = retained
`placenta; EFC = early first calving.
`
`It also is possible that the proliferation of indices
`could lead to the development of more strains within
`the Holstein breed, with greater inbreeding within
`each strain but greater diversity overall when crossing
`strains. Such schemes are similar to breeding strategies
`proposed for nucleus herd programs (e.g., Meuwissen,
`1998), although they are more common in the swine
`and poultry sectors. Some breeding companies offer
`mating schemes based on the assignment of young sires
`to genetic lines within the breed (e.g., Select Sires Inc.,
`2020), although the manner in which bulls are assigned
`to lines is not clear.
`
`What Will US Indices Look Like in the Future?
`
`In many ways, 2021 is a critical year in the evolution
`of US selection indices. The number of traits evaluated
`continues to increase, and is overwhelming to many,
`which suggests that new ways to group and express
`traits may be needed. The planned addition of feed
`saved to NM$ (Council on Dairy Cattle Breeding, 2020)
`will heighten tensions between economic value and trait
`reliability, and might make it impossible to continue
`publishing a single index for use by all breeds. If pro-
`ducers begin to question the credibility of the index,
`then they may turn to commercial indices developed by
`AI companies or, in the case of larger farms, develop
`their own custom tool for ranking animals for selection.
`Adoption of Subindices. As the number of traits
`included in selection indices continues to grow, there
`is growing interest in constructing NM$ as the sum of
`subindices, which Cole and VanRaden (2018) discussed
`in some detail. The idea is straightforward: instead of
`presenting an index composed of all traits of economic
`importance to farmers, biologically similar traits are
`grouped together into subindices that represent the
`expected portion of an individual’s NM$ due to their
`genetic merit for that set of traits. The total merit in-
`dex is thus the sum of the subindices. The principal
`
`advantage of this approach is that it permits breeders
`to focus on groups of traits that are important, such
`as cow health, without the need to focus on the indi-
`vidual details of each trait. For example, Table 3 shows
`a hypothetical version of NM$ that is constructed as
`the sum of 6 subindices: production, longevity, fertility,
`type, calving ability, and health. This example includes
`some traits that are not currently included in NM$
`(heifer livability and gestation length), and the weights
`are strictly speculative, but it demonstrates how the fo-
`cus shifts from individual traits to functionally similar
`groups of traits. Feed intake might initially be grouped
`with the production traits, but if additional efficiency
`traits are added, such as methane or carbon dioxide
`production, a sustainability subindex could be created.
`Although grouping traits into subindices would be a
`new approach for NM$, the Total Performance Index
`(Holstein Association USA, 2020) has been constructed
`using this strategy for many years.
`Can We Continue to Use the Same Index
`Across All Breeds? Historically, USDA has used the
`same selection index weights for all breeds. When a
`phenotype is not available for a breed, such as calving
`traits in Jersey, a value of 0 is assigned to that trait
`when the index is computed. This means that the rela-
`tive emphasis each trait receives in the index varies by
`breed (Table 4), which may not be obvious from the
`technical documentation (e.g., VanRaden et al., 2018).
`There is no reranking within breeds because all ani-
`mals receive the same value for the traits that are not
`evaluated, but as the differences between breeds grow,
`the index weights will no longer be optimal. In theory,
`there should be a breed-specific version of each index
`for each breed, which would result in 24 indices in place
`of the current 4.
`This problem is likely to increase in magnitude due to
`the planned addition of feed saved (FS) to the indices,
`which will have a large economic value (VanRaden et
`al., 2017) and relatively low reliability compared with
`
`Journal of Dairy Science Vol. 104 No. 5, 2021
`
`Exhibit 1039
`Select Sires, et al. v. ABS Global
`
`

`

`Cole et al.: INVITED REVIEW: SELECTION DECISIONS AND BREEDING PROGRAMS
`
`5118
`
`many other traits (Tempe