Comment by fbdab103 | Hacker News Reader

Comment by fbdab103 | Hacker News Reader

Selective breeding always struck me as interesting for the sheer amount of time required for experimentation. Pick your pairs, and then wait 2-5 years before you can repeat the cycle again.

Amazing that a single lifetime was able to yield measurable results.

Also reminded of the fox domestication program[0] where they were able to go from wild foxes to a dog-like domesticated breed within a decade.

>Belyaev was correct that selection on tameness alone leads to the emergence of traits in the domestication syndrome. In less than a decade, some of the domesticated foxes had floppy ears and curly tails (Fig. 2). Their stress hormone levels by generation 15 were about half the stress hormone (glucocorticoid) levels of wild foxes. Over generations, their adrenal gland became smaller and smaller. Serotonin levels also increased, producing “happier” animals. Over the course of the experiment, researchers also found the domesticated foxes displayed mottled “mutt-like” fur patterns, and they had more juvenilized facial features (shorter, rounder, more dog-like snouts) and body shapes (chunkier, rather than gracile limbs)

[0] https://evolution-outreach.biomedcentral.com/articles/10.118...

Think in scale. If you don't breed your cows, the neighbor's bull will come over the fence and do it for you, so you're incentivized to keep your cows bred. One bull will handle a large number of cows, so picking desirable bulls and a variety of interesting cows gives (roughly) $nCows potential picks for the next generation per herd (and Bakewell had many). Cows gestate for 8 months and results are frequently apparent in very young calves. Sexual maturity hits at 12-15 months (though, traditionally, first calf heifers were bred closer to 2 years), so 4 generations per decade in direct lineage isn't impossible.

It's not particularly difficult to select for specific traits when you consider inbreeding (which Bakewell did) and linebreeding (same idea, but avoiding excessive coefficient of inbreeding). Arguably, changing the muscle distribution on a domestic cow given a plethora of similar domestic breeds with different muscle distribution is easier than shrinking the adrenal gland in a wild fox given only wild stock.

Not to diminish in any way Bakewell's accomplishments. He revolutionized the field of animal husbandry.

Late edit to add an example: our (big, light brown, horned) Jersey milk cow came into heat and the (smaller, oreo-colored: almost black, white band around the belly, almost black; no horns) Belted Galloway bull jumped the fence, siring a heifer with small horns, a redder black, a splotchy band, and significantly greater size. Her offspring (against a bull from the same herd, possibly her sire) are medium-dark brown with a red tinge and a splotch of white, have no horns, and are still larger. Their offspring are very dark, solid brown with obvious red in sunlight, are still larger than the non-Jersey lines, and have, in every example, excellent conformation. All this in the last 7 years. The most recent generation will calf next year. Imagine what we could do with a targeted selection program!

I would think of it as annealing/gradient descent instead of some kind of Markov process where maybe you get lucky every 2-5 years. You are not breeding for the purpose of getting one animal with the trait (well you can, but not necessarily) so much as you are trying to influence a population of animals to display the traits you want over time. Maybe first generation you have an average yield of X across 100animals. And you don’t let the lowest yielding 50% breed (or just have them breed less so you can’t counteract inbreeding), now generation 2 has an average yield of 1.05X across 100animals. If you know the trait is heritable it seems pretty straightforward.

What’s challenging is I guess you can’t know a priori much about the actual genetics at play so at a certain point it becomes hard to know what’s simply phenotypic variance independent of genetics and what’s actually genes - it could take quite a while to be certain you’ve “converged”. Or for something more subjective/without an obvious strong genetic link, it could take a while to rule out the possibility of breeding for the trait at all.

Like for humans, I’m highly confident you could do some basic process over 10 generations to breed some ridiculously tall people. But it helps we know it has huge variance, is highly heritable, is easy to measure, and is polygenic. Probably a lot harder to try to breed “nice” people or people with thick skin. The fact dog breeders can do things like that is much more impressive IMO.

Wow, this is super interesting, thanks for sharing. A bit unclear from the articlr overview if they managed to do 6 reproduction cycles in just 6 years total and got to significant results. Or if it was 6x6. In any case, now my dream of breeding a fully domesticated group of raccoons seems more feasible!

Reply on news.ycombinator.com