Across this site I have taken the coat color genes apart one locus at a time: Extension, Agouti, the K locus, dilution, the brown locus. Breeders eventually ask the obvious next question. How do you put them back together to predict an actual litter? The honest answer is that you cannot read the loci in isolation and add them up. They are arranged in a hierarchy, where some switches override others, and predicting a puppy means working through that hierarchy in the right order. This article is the assembly manual.
Epistasis: why order beats addition
The key concept is epistasis, which simply means one gene masking the effect of another. Coat color is the textbook example. A dog can carry the genetic instructions for a beautiful black-and-tan pattern and still be born pure white, because a gene further up the chain switched the whole pattern off. The downstream genes are still present and will pass to puppies, they just cannot be seen. This is why a white shepherd is a genetic surprise package: it is hiding a full color genotype underneath a coat that shows none of it.
So prediction is not “average the parents.” It is a decision tree. You evaluate the loci in their hierarchical order, and the first one that says “stop” determines what you actually see, while everything below it goes along for the ride as hidden information.
The hierarchy, top to bottom
Extension (E locus, MC1R) comes first. This is the master switch for whether dark eumelanin pigment can be made in the coat at all. The recessive e/e genotype shuts off coat eumelanin entirely, producing the cream-to-white shepherd coat regardless of everything below it. I cover this in DNA testing at the E locus. The rule to memorize: e/e masks everything. If a dog is e/e, no other color gene will show in the coat. If the dog has at least one E, pigment can be made and you move down to the next switch.
The K locus comes next. This decides whether the Agouti patterns are even allowed to express. The dominant black allele (K^B) overrides Agouti completely, painting the dog solid dark; the non-dominant allele (k^y) lets Agouti through. This is the crucial ordering point I stress in the K locus and dominant black: K is read before A. A dog can be genetically black-and-tan at Agouti and never show it because K^B sat on top.
Agouti (A locus) comes third, but only gets a vote if the dog passed both prior gates (has an E, and is k^y/k^y). Agouti then sets the actual pattern: sable, black-and-tan, or recessive black, by controlling how eumelanin and phaeomelanin are distributed along each hair.
The modifiers come last. Once the base color and pattern are determined, dilution (D locus) can lighten black to blue and liver to isabella, and the brown locus (B) can turn black pigment to liver. These do not change the pattern, only its shade. They act on whatever the switches above produced.
A worked example: white crossed with colored
Take the mating breeders ask about most: a white (e/e) shepherd to a colored mate. Let the white dog be, as is common, hiding black-and-tan underneath. Suppose its full genotype is e/e at Extension, k^y/k^y at K, and a^t/a^t (black-and-tan) at Agouti. Its colored partner is E/e, K^B/k^y, and a^w/a^t.
Work Extension first. White is e/e, partner is E/e. The cross E/e × e/e gives half E/e and half e/e. So about half the litter will be colored and half will be white, on average. That alone surprises people who expected an all-colored or all-white litter.
Now look only at the colored (E/_) half, because the white half shows nothing regardless. At K, K^B/k^y × k^y/k^y gives half K^B (solid dark) and half k^y/k^y (pattern shows). So among colored puppies, roughly half are solid black-derived and half display an Agouti pattern. Among the pattern-showing ones, the Agouti cross a^t/a^t × a^w/a^t yields a mix of sable (a^w) and black-and-tan (a^t). The white puppies, meanwhile, silently carry their own combinations to pass on later.
That is the whole method: never blend, always cascade. Resolve the top switch, carry the survivors to the next, and stop where masking occurs.
A note on color calculators
Online coat-color calculators are popular, and they are genuinely useful, but only if you feed them the right input. A calculator is just this cascade automated: you enter each parent’s genotype at every locus and it runs the Punnett combinations and applies the masking rules for you. The arithmetic is where calculators shine, because tracking five or six loci across a litter by hand is error-prone.
The catch is that a calculator can only work with genotypes, not guesses. If you type in a white dog as “white” without knowing what it hides at K and A, the tool has nothing to cascade and the output is meaningless. Garbage in, garbage out applies with full force here. So the calculator does not replace DNA testing; it consumes its results. Test the parents at the loci you cannot see, enter the verified genotypes, and the predicted percentages become trustworthy. Skip the testing and even the best calculator is just guessing in a prettier interface.
How hidden recessives surface
This cascade explains the “impossible” puppies. A breeder pairs two solid black dogs and gets a black-and-tan, because both parents were K^B/k^y hiding a^t underneath, and a k^y/k^y puppy finally let Agouti speak. Or two colored dogs produce a white puppy, because both carried a hidden e. The trait did not appear from nowhere; it was masked in the parents and merely uncovered in the offspring.
The practical lesson for litter planning is to test the loci you cannot see rather than trusting the visible coat. A colored dog’s phenotype tells you what is on top, not what is hidden below it, and the hidden alleles are exactly what determine your surprises. Combine that genotype knowledge with sound genetic diversity management and you can predict not just what a single litter will look like, but what your line will keep producing for generations. Color genetics rewards breeders who read the switches in order and write the hidden alleles down.