Contention point #17:  "Inbreeding" is strictly a "pedigree" issue

Back in the days before DNA markers, bloodtyping was used to verify parentage on embryo transfer calves.   Paclamar Farms of Colorado (breeders of "Astronaut", "Bootmaker", "Triune Complete" etc) was attempting to do flushes mixing semen on two sires.   They also were followers of the Weeks Analysis ("aAa") breeding guide, in planning their matings.   It became impossible for Paclamar to do mixed flushes, as the sire pairs they would propose on their donor cows could never be sorted on bloodtypes.   This was true even when the sires they were proposing had no close pedigree ancestors in common.   Why would this be true?   Simply, if animals have a similar phenotype (on which "aAa" ratings are based), they received that from a similar gene inheritance.   The phenotype of the animal reflects the genotype received from conception.

If you look at a pedigree, what does it tell you?   Who the ancestors were.   Does it tell you which of the ancestral genes were passed to the animal in question?   No.   We can only determine that from an expensive DNA scanning, or from more inexpensive phenotypic analyzation, such as the Weeks (aAa) people do, and from testing, classifying, and maintaining health and reproductive records.   To put it in another way, I will quote the late Dr Meadows of MSU (Kellogg Farms Guernseys): "the genetic inheritance of your animal comes from two ancestors: the sire and the dam.   Nobody else in the pedigree matters."   Biologically, this is 100% true.   Geneticists sometimes lack biological clarity when it comes to expressing their ideas by mathematical equations, and so they tend to lead us into mental minefields, such as "inbreeding coefficients" and "expected future inbreeding" ratios.   These expressions lead us to the belief we might need to look at three, five, maybe eight generations of a pedigree to be sure we have all possible "inbreeding" accounted for in a mating.   This is in reality only necessary in seeking the possibility of latent lethal "recessive" genes that can hide for generations of matings to non-carriers..

I think the Dutch found a much simpler explanation for the bad gene effects we are calling "inbreeding depression".   They looked at the results of three generations of cows produced from using the highest Protein index sires obtained from North America on their (99% unrelated) native Friesian cows, and define all the bad stuff that was happening (lower fertility, more calving difficulty, shorter herdlife, more mastitis, more veterinary expense) as "selection depression"-- the inadvertent losses you get when you are basically following a single-trait selection scheme (PTA Protein yield is all that matters).   They got what they selected for-- big gains in Protein yields-- they also got what they weren't paying attention to-- less fertility, less disease resistance, less calf vigor, shorter herdlife).

Back in the 1930s, 1940s, and 1950s, some great Breeding herds inbred themselves out of existence by the effects of close matings (sire to daughter, son to dam, brother to sister) over many generations.  Genetic recessives (like "Blad" and "CVM" in Holsteins, "LL" and "RVC" in Jerseys) often come from lines that were either inbred or highly linebred.   The point is, when you pick the ancestors for your future herd, it is highly impractical to pick them in a "single trait selection" sort of way (as Brampton did in focusing upon Brampton Basilua because of a 1313 pound world record butterfat production).   The cow and/or bull you pick has the "strengths" you see, but may also have some weaknesses you don't see (or choose to ignore as not a problem in your existing base herd).    Close breeding intensified the potential inheritance for the trait you wanted-- it also intensifies any weaknesses.   At some point the weaknesses will overtake the strengths, and overnight-- in a single generation-- all you gained can be lost.   This was identified as "inbreeding depression" in the 1960s after study of these famous herds.    But the success of other closely linebred bloodlines (such as Canadian Rag Apples) over decades was not thought to be a contradiction of the research conclusion.

What we are doing today in AI-- mixing all our sire lines together in our bull and cow pedigrees-- is not "inbreeding".   It is more the sort of "hybridizing" that is done by corn seed companies to produce a high yielding "terminal cross".   The problem: hybrids do not breed true to any trait pattern.   It does require some level of linebreeding, as plant breeders and poultry and hog breeders do in their proprietary breeding programs, to preserve an intensity of trait transmission.   They then cross these lines together to produce the production plant or animal that performs well in the commercial field.   We are not maintaining linebred dairy seedstock lines in the same way that successful poultry, swine, beef and plant breeders have done, and that has its influence upon our trait heritability levels overall.

Conclusions on solving inbreeding

Dr Ben McDaniels, in his series of articles detailing the inbreeding question for Jersey Journal, came to a conclusion that we should be maintaining a variety of sire lines-- at least three lines minimum-- and that these sire lines should be linebred within themselves.   Then in our national Jersey herd, we can rotate these sire lines through matings such that the inbreeding levels in our herds would not increase, thereby avoiding the risks of "inbreeding depression".    Dr McDaniels has clarity on the biology.   To date, no AI systems have seriously attempted to implement his cogent recommendations. 

Contention point #18:  "Reliability" really means something

There was a point in the late 1970s /early 1980s when large numbers of new sires were coming out of the neighborhood sampling efforts of large California dairies.    These bulls had high lactation averages, and high PDs, on their west-coast offspring, and were leased by major AI systems.   We would often see these summaries collapse after use across the country, in all management regions.   The answer to all this was to demand higher "Repeatability" (more herds, more regions) in samplings to guarantee a higher level of statistical accuracy.

It has been interesting to travel the USA west coast, observe the technological marvels of large-herd monocultural dairying, and discover that today's west-coast bull has a lot less trouble with "Animal Model" than the previous modified-contemporary evaluations.   Of course, with "Animal Model" you no longer have just "Repeatability" (standardized number of daughters per herd, maximum number of herds)-- you have "Reliability", a term unique to "Animal Model's" inclusion of siblings and parents into the evaluation calculation.

You see, what I was learning, is that many bulls who started out successful on the west coast, remain successful into second generation usage on the west coast.   This suggests the first-generation data on those older west coast bulls was probably accurate-- for that region.   In the 70s and 80s, dairying in the midwest and east coast had not yet adopted all the west coast characteristics (group corrals, bunk feeding, total mixed ration, commodity feeds, higher frequency of parlor milking).   Today we see the west coast dairy model being copied everywhere for expansion dairies.   "Regional effects" become less distinct (pretty much limited to weather and soil influences on crops raised for cows).   Genetic evaluations, therefore, are more "reliable" when coming from regional sampling than was previously true.

Selection for higher Reliability, therefore, emphasizes not "larger samplings make for more accurate data", but "sticking to the most familiar families as sources of sampling sires guarantees a higher Rel% from sibling and ancestor merits".   Today, a bull from a familiar family can hit 85% Reliability in an organized sampling program with 20 fewer daughters than an outcross bull sampled outside of such captive herd systems.   This in fact makes it easier for the private syndicate to "fake" a proof than was ever possible before, and they don't even know they did it, because they (1) sampled in lots of herds, (2) gave no preferential treatment.   The faking comes from the preferential treatment (on both the managerial and evaluational level) given to the high indexing cow family with many ET flush members.   (For the Holstein people with long memories, think Eric Dew Quietcove Michael and Singing Brook NB Mascot, both high Rel% AI sampled sires.)

Contention point #19: define "breed average" around regional effects

Trouble is, over time these captive herd (contracted) sampling environments, which use as high as 50% young sires to receive sampling incentive payments, become skewed from the attained breed average within the AI market focused upon using the best evaluated sires.   If half your heifers are  sired by sampling sires (90%+ who will never enter active AI service following first evaluation), over time the "contemporary valuation" will be based as much upon "pedigree merit" as "relative values".  This has the effect of padding the range of PTA values upward of reality.  

Because contemporary sires (sampled to 85% Rel) never receive future AI daughters, their genetic "value" is fixed within a contrived environment.   By comparison, sires who "graduate" to AI service receive future offspring, who over later generations of matings in the "real world" against 99% Rel contemporary sires, change their evaluations.   As most AI sires are overevaluated (as a consequence of using "Animal Model" within the contrived sequestration of contract sampling herds), we see more sires go down than up-- especially for type ratings, where sampling herds deviate most widely from the purebred herd average.   Therefore, they do not survive that second generation usage.  

Why did all the AI systems in the USA end up going to the contract sampling herd system?   Because the studs who did it first were getting higher "proofs" than the studs who stuck to open usage sampling.    This does not mean they are generating better bulls.   By the time we know for sure, the bulls are old, and we're looking for something new instead, so very few people notice.

Contention point #20:  genetic progress is measured by the pedigree indexes on your heifers

Thinking of the points raised above, let us assume your herd is suffering a 40% annual cull rate.   This basically means, if you are raising 100% of the heifer calves born, you also need 100% of them just to maintain herd size.   We already know from the "bell curve" theories that not every one of those new heifers will be an improvement over her dam (at least, when using the politically correct "random mating").   Therefore, all possible genetic progress from generational turnover is sire dependent.  

If at the same time, we later find that a large percentage of the sires we were using based upon first generation evaluations, cannot survive a second generation of usage (ie, were overevaluated), does it not seem logical to then theorize our "genetic progress" has been stymied by (1) not having a surplus of replacements to sort "genetically", (2) having too low a "harvest rate" on newer AI sampled sires?

The initial intent of genetic ranking composites was to design a system that mimics the economics of dairying.   We would assign "values" based upon our income sources (milk components, quality) and costs of production (replacement raising, feed cost, labor, vet care, reproduction cost).   Each bull would be "ranked" based upon how he does as a total package of evaluated traits.   It's a great concept.   It also does not appear to work    Cows have lower fertility, more veterinary contact, and shorter herdlife than their ancestors.   This suggests the various genetic trends are neutralizing the potential for overall {economic] genetic progress for large numbers of dairymen.

To support this argument, I would suggest a review of the "Top 100 TPI sires at 99% Rel" from any recent year.   Within this list, check for how many of these sires never spent any time on the "Top 100 TPI" list as typically published.   The significance of this is that they displaced sires who had a head start by being ranked from their first (sampling) generation evaluations.   Not only do lots of bulls go down, but they go down far enough for a lot of decent non-ranking sires to take their place.   Of course, the irony is that those underranked sires that rise to the top are then considered "too old" to contribute sons for AI sampling, leaving us with many sons of overranked sires, who are also unlikely to add to genetic progress.    "Indexing" biology is a two-edged sword, and it appears both sides of that blade can injure us.. primarily by reducing the pedigree variety of surviving sire lines .               

Conclusions and suggestions

If there is any validity to ranking of sires, this ranking should be within subgroups only-- for instance, rank all the "Rudolph" sons against each other-- but otherwise, rank sires according to individual traits only, and if there is any validity to the concept of "composite indexes" rank those only according to a composite of individual trait percentile rankings.

The benign belief that current generation superior sires are so good they can be used randomly (ie, blindly as to the trait needs of the cow being bred) has, in the opinion of many, become a malignancy in terms of the losses in fertility rate, veterinary expense, and herdlife being seen in our high production herds.     The "genetic effect" from selection of high performance ancestors can be neutralized by the "mating effect" of picking a mating sire who cannot correct the faults of the mated cow.    It is a cow's "faults" that will cull her from productivity, either now or later in life.   Extending those faults into her offspring to a greater degree (or adding faults not already possessed by the cow) produces new cows who are culled at even earlier ages.   The trait similarity in ranking sires accelerates the rate of negative trait accumulation just as fast as the theoretical acceleration of "genetic value" for the priority focus selection traits.  

"Composite Index Ranking" promotes the biologically indefensible concept that "perfect" bulls exist.    This more than anything is what should reinforce a recognition that population genetics is mired in the theoretical realm of mathematical constructs.   Over time, first peripheral and then even basic trait patterns will change as a result of genetic selection.   (If you look at who the "ranking" Holstein sires were in 1964, none of them have a surviving sire downline today.)    Economic changes dictate adaptation changes in our animals, both in physique and performance characteristics; future adaptation requires we preserve [lower ranking] lines that possess trait variety that may be needed as the environment demands them.   (Two Holstein examples: Osborndale Ivanhoe and Romandale Reflection Marquis were moderately-ranked sires that filled the demand for a faster-growing, taller, higher uddered, fast milkout cow in the 1970s switch to parlor milking.)

The majority of cows are either culled today for mastitis, or physical defects leading to declines in production, or infertility, or lost from post-calving metabolic disease.    These are all areas that are genetically influenced; the sum total of these adds up to their potential for productive life.   We now have genetic indices for mastitis (somatic cell score), some physical traits (linear evaluation), cow fertility (daughter pregnancy rate), and we also have the overall measure of survivability (productive life).     What we do not have however is any consensus among geneticists or sire analysts that these measures should influence selection of sires for future AI sampling.    If our sire offerings therefore have been selected in the absence of consideration of traits having the major impact upon cow herdlife, our selection methods for the sires we use have to compensate for this defect.    By this logic, mating cows individually to compensate for her inherent faults, to sires known to be superior specifically for these faults, has a better chance of being successful in reducing future rates of herd turnover.   Only with a decrease in involuntary culling can we actually have the surplus of productive cows available to actually allow us to "cull for genetic progress"-- and there is no evidence available that selection by composite index alone has achieved that within any high production environment.