Thread by @thattai

In dogs, three genes *independently* determine if the coat is short or long, soft or wiry, and straight or curly! science.sciencemag.org/content/326/59…

This kind of "gene-for-X" story is actually very rare, despite what the news headlines often claim. So why is it true in this case?

Knowing how cells work, it's not surprising most traits are "complex": they depend on 1000s of genes:
quantamagazine.org/omnigenic-mode…. Conversely, mutations in most genes are "pleiotropic": they affect 1000s of traits.

So when would we expect a single gene to control a single trait?

Genes encode both "material" and "regulatory" proteins (as well as regulatory RNA and DNA elements).

Just like a tap controls the outflow of hot and cold water, regulatory proteins control the levels of the material proteins that give a cell its physical and chemical properties.

Across dog breeds, or in closely-related species like humans and chimps, material proteins are nearly identical. But changes in regulatory proteins, RNA and DNA elements, allow these material proteins to be remixed to give each breed or species its traits. cell.com/fulltext/S0092…

So how best to regulate the mixing of material proteins? Think of how we do this for hot and cold water.

System A: Independently control each separate input (hot, cold).

System B: Independently control some properties of the output (temp, flow).

Each has its own advantages.

Any temperature and flow rate can be achieved by both [h,c] and [T,F] controls. The systems are thus mathematically equivalent, but have different engineering advantages.

If the "operating point" is fixed, you would probably use the [h,c] system since it is mechanically simpler.

But what if we want to change properties of the output? E.g. keep T fixed while raising F.

The [h,c] controls need TWO adjustments. With [T,F] controls we need just ONE.

Also, wobbles in the h (or c) knob mess up BOTH T and F. Wobbles in the T (or F) knob mess up only ONE.

Mutations in a gene encoding a material protein, or mutations in a regulatory gene of the [h,c] type, will affect many traits.

Mutations in a regulatory gene of the [T,F] type will affect a single trait, leaving others unchanged.

Where would we expect to see [T,F]-type genes?

We've seen that [T,F] controls are useful when the operating point is changed.

If a population is put in a new environment, individuals with mutations that affect just one fitness-enhancing trait are more likely to be selected, compared to individuals with pleiotropic mutations.

Whether such mutations exist depends on luck. But *if* they do, individuals that harbor them will soon sweep the population.

So [T,F]-type regulatory genes are evidence of recent selective pressure. In dogs, this pressure was applied by breeders, to produce interesting coats!

I discuss some of these ideas in my talk on "The Unreasonable Irrelevance of Mathematics to Biology": online.itp.ucsb.edu/online/bblunch…

I'll just close by saying: to deeply understand living systems, we are often required to pull together ideas from cell and molecular biology, developmental biology, genetics, evolution, engineering, physics, and yes, mathematics.

That's what makes it so much fun.

/fin