Refining the "genetic" difference: herdmates, contemporaries, parity groups, test day

In the 1970s, we see the inauguration of the "modified contemporary comparison", which was a first attempt to "index" the "quality" of herdmates, and a recognition of the need for generational "parity".   Two changes were made: (1) the herdmate average was "adjusted' according to the average PD of their sires; (2) two "parity" groups were defined, first lactation cows, and later lactation cows.

The point of the (1st) change was to suggest if a bull was sampled in a comparison against daughters of "minus" bulls, he had a better chance of making a big "plus" over herdmates than if he was sampled within herds using only "plus" sires.   The point of the (2nd) change was to compare first-lactation daughters only against other first-lactation cow herdmates, and daughters with later lactations only against other cows' later lactations.   This avoided the possible intergenerational biases within herds where cows lived full lifetimes, allowing for management improvements to "bias" the comparisons.   In this era is where the concept of "genetic trend" [ie, the newer generation is always better] first made its debut in dairy genetics' discussions.   A higher weighting was placed upon first lactation comparisons than later lactations, to give the faster maturing animal an advantage on PD calculation.

In the 1980s, greater computing power made it possible to do full justice to age "parity".   In this time we moved to a contemporary definition by age groups.   Roughly, you had five overall "parity" groups to define a contemporary: 1st, 2nd, 3rd, 4th and 5th lactations.   Later lactations were no longer used for sire or cow indexes (ie, 7 yr old and later lactations).   The relative weights for the first five records were as follows:  1st 78%, 2nd 14%, 3rd 4%, 4th and 5th 2%.   In other words, while we went to a deal of trouble to define a herdmate accurately (as to being unrelated by sire and at the same age) we also made a philosophical decision to define first lactations (and for type, first scores) as the most "accurate" lactation to define genetic differences.   The assumption: later age performance is subject to more "management intervention"-- plus, larger percentages of contemporaries have been culled so we are left only with a comparison of the superior against the superior (ie, less measurable differences).

Before you question the influence of breeding philosophy on genetic evaluation, consider that under current "Test Day Model" evaluation as practiced by Canadian Dairy Network, the first three lactations (only) are used for PTA calculation, but the weighting is equal between 1st, 2nd and 3rd lactations.   The Canadians note they saw reranking of sires as a result of this change, as well as more "stability" to the ranking over time (ie, less PTA fluctuation from later addition of evaluated offspring).

The big change made in the later 1980s, was adoption of the "Animal Model" calculation.   This was to replace the "PD" [Predicted Difference"] with the "PTA" ["Predicted Transmitting Ability"].   Under Animal Model, we make a major shift in assumptions-- namely, that higher herd averages indicate higher genetic ability environments.   The daughter average is now compared, not just against a parity of contemporaries, but with those contemporaries' overall management level factored in.   But the biggest change under "Animal Model" is to begin calculating PTAs from a "Parent Average" and to include within the PTA as additional information, all sibling PTA values weighted by closeness of the relationship to the animal evaluated.   A PTA value, therefore, begins from Parent Average, adds in sibling averages next, then adds in progeny last (as more progeny are added, the pedigree weighting is lowered).    

The "Animal Model" is still in use today.   The only change has come from the 2000s adoption by some countries of "test day model" which means, instead of 305-day lactation comparisons, we have a 30-day "lactation" (ie, each test day) comparison.   This supposedly raises accuracy by lowering the impact of the occasional off test day (cow was injured, in heat, had mastitis, etc on test day)-- plus it allows for some additional calculation such as "lactation persistency" comparisons.   Many countries adopting test day model have gone to a three lactation parity-- using 1st, 2nd and 3rd lactations, and reweighting them according to that countries' belief in which lactations offer the most accuracy for evaluation comparisons.

 Contention point #1:  Mature Equivalent factors

Mature Equivalent factors were developed in order to make daughter-dam comparisons valid in an era when it was recognized cows could produce up to 50% more at "maturity" (5 to 8 yrs of age) than they did on first lactations (2 yrs of age).   To compare immature daughters' lactations against dams with mature lactations, the ME factors were developed.

Today, with all calculations being done at "parity" age, why do we still use ME factors?   These add 30% to 40% to first lactation actual yields, thereby they make the range of PTA values 30% to 40% larger than the actual differences between new bulls.   Given a national average of 2.4 lactations per cow on DHIA, less than half of all heifers freshened ever reach a "mature" lactation age; the use of ME factors gives perhaps two thirds of our cows (and their sires) a PTA advantage they will never yield in actual pounds of salable milk to dairymen.

Our friend Jack White of Valleystream Jerseys calls "ME" records "Maybe Eventually".   We think he hits the genetic nail right on its economic head.   We only get to sell actual pounds of milk.

Contention point #2:  Parent Averages dictating base PTA values

The decision to include pedigree/parent averages within PTA calculations results from a genetic study of 1970s sire summaries, in which it could be shown there was a general correlation between pedigree index at sampling and resulting PD values at 80%+ Repeatability.   The weakness within this was that nobody tracked sire lines beyond a generation or two to see if they stayed ranking performers.   In fact, when you look at the 1964 sire summary as an example, and identify the top 10% of PD sires at that time, you find none of them have a "downline" remaining as top PD sires by the 1980s.   To use some prominent examples in today's high PTA Holstein sire lines, within three sire generations of the sire defining the line, you will find a minus PD grandsire** in direct line:

Carlin M Ivanhoe Bell -- Penstate Ivanhoe Star -- Osborndale Ivanhoe -- Osborndale Ty Vic**   Pawnee Farm Arlinda Chief -- Pawnee Farm Reflection Admiral -- Rosafe Pearl Hannibal**        Round Oak RA Elevation -- Tidy Burke Elevation -- Wis Burke Ideal -- Wis Ideal**                  

Breed histories are full of unique sires who changed the direction of breeding, whose full or maternal brothers were relatively worthless.   In Jerseys we could suggest Milestones Generator, who had four full brothers used in AI and all ended up "minus", yet he was himself a huge "plus" and started a sireline of his own (Generator HL Earl -- Bold Venture -- MVF Bold V Daniel) that endures.

One problem is that we are confusing the "heterosis" effect with the "genetic" effect.   You can have heterosis response (outcross sire onto linebred cows) just as easily within a breed as when cross breeding.   For example, Madawaska Aerostar (double grandson of "Elevation") was an outcross to both "Arlinda Chief" and "Bell", and his performance in the USA exceeded his "Parent Average" by 300% in contemporaneous use.   Likewise his sons have been remarkably successful wordwide,  

I really do not see the problem with returning to a "base zero pedigree" PTA calculation.   We identified the genetic superiority of the breeds' sireline progenitors (Ayrshire: Bettys Commander) (Brown Swiss: Elegant) (Guernsey: Minnies Choice) (Holstein: Arlinda Chief, Elevation) (Jersey: Chocolate Soldier, Quicksilver) under a system that waited for progeny data before it calculated any evaluations.   The inclusion of pedigree ancestral and sibling PTAs within sire (and cow) evaluations delays, rather than advances, our recognition of the real transmitting value of today's sires, especially those who could offer outcross heterosis..  

Contention point #3:  Sibling PTAs having equal weight to progeny data

There have been several genetic studies tracking the genetic rankings of ET siblings which show the old "bell curve" of response holds pretty true.   The only thing identical about full siblings that are not split-embryo twins, is their parentage.   The number of gene combinations possible from any mating of two unrelated mates is very high, and therefore the probability of identical performance is very low.   Yet under "Animal Model" we postulate the performance of both paternal and maternal half-siblings to have a direct bearing upon the genetic value of the animal under evaluation.

The first purebred Holstein cow to have over 50 registered offspring was Diamond S Eve Corry (EX) who was a sixth generation scored "EX" in a maternal line.   She produced a daughter whose female descendants have now stretched this chain to over ten "EX" generations, and she had a son, Carnation Counselor, who at one time was the highest type-rated sire in the Canadian sire summary; yet among her 50+ registered ET offspring can be found five daughters who scored below 60 points for type.   No amount of pedigree depth can overcome the occasional mispairing of genes that occurs during the moment of conception, and it is ludicrous to suggest a 52-point daughter of "Corry" has a greater ability to transmit "type" just because she has a maternal half-brother named "Counselor"!

You will see sampling pedigrees on today's leading indexed young sires, with Rel%s on their PAs already at levels of 55% to 60% in some cases.   This is the result of having lots of ET flushmate siblings included into your PTA, at the day you are born.   This sort of young sire can reach 80% Rel on an evaluation with a minimum number of (preselected) offspring, but it does not make him any more reliable in transmitting ability when you recognize that those siblings all represent genetic combinations that differ (from point of conception) from the sire's personally unique gene makeup.

Contention point #4:  The first lactation has the most genetic significance  

I think cows are like cars.   Some are designed to only last 60,000 miles (ie, as soon as the warranty runs out), others are engineered to last 300,000 miles.    Both of them, with high octane fuel, could hit 60 miles per hour within ten seconds, but what defines the difference in their value is not acceleration, but durability at highway speeds.

Geneticists recognize the greatest difference between cows in productive ability is in first lactations, because (as learned from the raw data used to develop ME tables) not all cows mature at the same age or rate.   Production volume is greatest at "maturity", therefore the earlier maturity can be reached, the greater will be early lactation production volume.   Our sire indexing for "more milk" has actually had the side effect of selecting for "earlier milk" (culling the bloodlines who have the longest delay to, and yield improvement at, maturity).

If cows could be raised for free, there would never be any problem with first lactation selection.   As the two-year rearing span for dairy cows represents a significant investment, there is an economic need for cows to remain productive over a multi-lactation lifetime.   Furthermore, for any possible future genetic progress, cows need to provide their own replacement heifer ready to milk by the time they are "spent".    Economically, it would be better to base genetic measurement upon pounds per day of life, or even better, first three lactations total production.   This would reduce the current advantage slow fertility/short herdlife cows have in the evaluation system.  

Contention point #5:   Genetic progress is measured by higher heifer than cow PAs 

Genetic progress can only be made if we have sufficient herdlife to allow for culling on "genetic potential".   This means we have to generate a surplus of replacement females, beyond what is needed simply for replacment of involuntary culls.

Simple math tells us, with a 52-48% male-female calf ratio, for each cow to have a live heifer calf to replace her, we need each cow to calve twice.   If 24 months is average age of first calving, and we hit an optimal annual calving interval, the "average age at first calving with a heifer calf" becomes 30 months (24+36 divided by 2 calvings).   If we then assume 24 months to grow that calf into a freshened replacement, the average cow has to reach a minimum 54 months of age before she has a first daughter ready to replace her.   This is the same as a minimum 3.5 lactations average herdlife: 24 months old first calf, 36 months second calf, 48 months third calf, 6 months into lactation = 54 months attained age.   In the USA we are averaging 2.4 lactations in DHI herds, which means we have to import 60,000 cows per year from Canada, and milk most cows for longer than a 305 day lactation, just to maintain national herd size.    

On that basis, our "genetic progress" is being achieved on paper only, the result of an every-five-year rolling of the "genetic base" national herd average, against which sires are compared.   

Contention point #6:   Longevity begins at birth (not at beginning of lactation) 

It is great to send out 1200 straws of semen on a young sire, get 200 pregnancies, have 100 heifers born, of which 80 make it into his evaluation giving him a high Rel% sampling.   However, if we send out 1200 straws, only get 150 pregnancies, have 75 heifers born of which only 50 make it into his evaluation, and he generates higher PTA values than the bull with 80 daughters, is he truly the best bull to use?    (1) Conception rate was lower, (2) abortion and stillbirth rate was higher, (3) death and infertility loss of weaned heifers was higher.   None of these sound good for dairy economics, but we ignore all of that and concentrate on the ME lactation differences to define "genetic value".    As genetic fertility (highly correlated to longevity) has declined, AI systems have just raised the volume of sampling semen used to ensure a "high Rel%" proof.   We have invested the term "Reliability" with a biological meaning it can never possess.   There is nothing "reliable" about a bull who takes 24 straws of semen to produce one milking first lactation daughter, compared to a bull who can do the same job on only 15 straws of semen.   Add up all the factors, and consider whether their cost can be covered by a few hundred pounds of PTA milk difference between two sires?.   

Contention point #7:   305d Yield volume is more important than calving interval

We visited a farm recently who had trophies on the wall for the county high herd average.   At the same time his vet was telling him to give up on AI and get some bulls in his herd, because the average of first heats observed was over 100 days, and the average days open was over 240 days.   This is a herd owner who always followed the high index selection route, and what he got along with that high "Net Merit" selection was a genetically infertile herd.   Proof: after two years of multiple bull natural service, his average days open is "down" to almost 200 days.   Obviously, heat detection and skill of insemination was not the problem.

NCSU research has shown that there is a sireline difference in cow fertility levels, and that this has generally correlated with high milk yield sire lines.   The difference is -400 PTAM on first lactation cows and -1000 PTAM on later lactations.   In other words, measuring sire performance on the first 305 days of lactation milk yield, without consideration of calving interval, has given low fertility sires an unfair advantage over sires who maintain desirable fertility.

The two big "milk" sire line sires in Holsteins are "Ivanhoe Bell" and "Arlinda Chief".   Looking at the dams of these famous progenitors, we find the dam of "Chief" made five lactations averaging a 433 day calving interval; we find the dam of "Bell" made four lactations averaging a 506 day calving interval.   Linebreed these two for a few generations to get more "milk", and how much "fertility" do we get at the same time??   

Contention point #8:   Higher herd averages mean higher genetic value

Is it not ironic that back in the 1950s we switched to daughter-dam comparisons, recognizing that 80% of a herd average was feeding and cow comfort, and only 20% (the in herd variation) actually genetic?    Then in the 1960s reinforced this recognition by swiching to an ingeneration comparison against herdmates, recognizing that progressive (AI user) dairymen learn how to make their cows milk more, trending herd averages upward?    Yet we get to the 1980s and suddenly a high herd average out west (western irrigated hay, commodity TMRs) means those cows are genetically better than a low herd average out east (hilly pastures, grain at milking, winter silage), on identical sire stacks?   Relative value in herd (after all involuntary culling reasons) still determines which cows stay and which leave.   This is true whether your herd is a 15,000m grazing herd or a 24,000m confinement herd.  That Jersey cow who recently made over 40,000m third lactation on  rBst and oxytocin in a TMR fed herd is still the same cow who made 7,500m first lactation in an organic grass dairy.   The same cow genetically at five years that she was at two years.    Environment makes all the differences in the volume of yield.   Relative value in herd is the only "genetic" yield measurement.

Contention point #9:   All we need is the type score closest to 30 months of age  

When it comes to type evaluations, even though the option exists to reclassify any cow each lactation as long as she is in the herd, the only score that counts for genetic evaluation purposes is that first score.   Why?   Because all cows calve for the first time.   Beyond that point, it is "management" that determines whether a cow gets to calve for later calvings.   Therefore, second and later scores will contain too much "management bias" to be genetically valid.

Holy Cow!   Let us think about the reasons cows leave herds: (1) they get mastitis, (2) they do not get bred back, (3) they get injured during calving, (4) they die after calving, ie milk fever, ketosis/DA, or coliform mastitis, (5) they get lame, ie heel warts, foot rot, laminitis, (6) they dump their udder after calving, (7) they die of pneumonia, (8) they do not milk enough.   At least half of these could be the result of a dysfunctional physical structure (ie, "type").    As dead cows do not get rescored, I would agree that the "genetic clues to dysfunction" need to be picked up on the first score.   However, in reality, they do not, and as a result we see bulls +2.00 PDT who are -1.70 PL.

When linear type systems were inaugurated, it was the early 1970s, and what little research had been done on the correlations between type and production were done in stancion barn environments (ie, carry the feed to the cow).    The major conclusion drawn: highly angular young cows milk more as young cows.   This was enough to design a type system that puts 70% of its weight on feet & leg and udder traits, and the other 30% on just basically does the cow look like she is working hard?   As a result of this type framework, we now call cows "dairy" that our grandfathers would call "frail", as in "not strong enough to survive how much she milks".    Recent Penn State University research confirms that the "angular" young cow has higher SCC scores and requires more veterinary care.

What Breeders have noted is that there are sires whose daughters routinely "dump" their udders after a second or third calving (ie, before truly reaching maturity), whose type rating remain very high, so they wonder "Is it just me?   Am I just bad at making matings??"   In fact, the culprit is only using the first (ie, immature heifer) calving classification data.   If I wanted to make someone mad, I would start naming bulls who do this, but your own data (if you have been using only popular AI sires) will answer the question for you.

Prior to this assumption, which allowed a lot of sires to become "plus" for linear type (who started out "minus" under earlier systems), we had the Cornell University theory of "stayability"-- high PD milk sires' daughters survive longer than minus PD milk sires' daughters.   This was done by observing if a cow was still in milk at 42 months of age.   In the 1970s, average first calving age was 27 months, then figure 13 month calving interval, means 40 months at second calving, so at "42 months of age" a sire could get credit for "stayability" if his blown-uddered second-calf daughter was still being milked (ie, get her first 90 days flush production, then ship her).   In other words, type had no impact upon a sire's stayability, therefore type was a meaningless measure...   (Current genetic biases tend to feed upon previous great genetic ideas!)