The Gap Between a Sequence and a Trait

When the Human Genome Project reported its first draft in 2001, one of the more startling numbers was the count. The expectation had been something like a hundred thousand protein-coding genes. The answer came in at around twenty thousand, roughly what a nematode worm has, and fewer than a grape.

The surprise says more about the expectation than about the finding. Everyone had assumed complexity required proportionally more parts, and what the low count showed is that complexity lives somewhere else: in regulation, in combination, in when and where things are switched on. Alternative splicing lets one gene produce many different proteins. The regulatory sequences that determine expression outnumber and outwork the coding ones. The difference between a human and a worm turns out to be less about the parts list than about the control system.

Which is the theme of this chapter, because nearly every popular misunderstanding of genetics lives in the gap between a sequence and a trait.

Start with what a protein-coding gene actually specifies: a sequence of amino acids. That chain folds into a protein, and the protein does something: catalyses a reaction, binds a molecule, forms a structure. What it does depends on which cell it is in, what else is present, what signals have arrived, and when it was made. There is no gene for a nose. There is a gene for a protein that, in a particular developmental context alongside thousands of others, participates in a cascade that ends with a face.

This is why "blueprint" is the wrong metaphor and has done real damage. A blueprint maps one-to-one onto the structure: this line is that wall, and you can read the building off the plan. A genome does not work that way. The better analogy, imperfect in its own ways, is a recipe. The outcome depends on the process as much as the list, no single word corresponds to any part of the cake, and you cannot reconstruct the instructions by examining the result.

Next, the phrase "a gene for". It is not simply sloppy; it means something specific, and the something is always about a difference. To say there is a gene for blue eyes is to say that a particular variant is associated with a difference in eye colour, in a given population, in the environments that population inhabits. There is no gene for having two legs, because there is no variation in the trait to associate anything with. And the association is population-relative: variants associated with height differences among Europeans may say nothing about a population with different variation or different nutrition.

Most traits of interest involve thousands of variants, each contributing an amount too small to detect individually. The headline "scientists find the gene for X" is almost always describing one variant of tiny effect among many, or a rare mutation with a large effect in a handful of families, which is a genuinely different kind of finding.

Heritability compounds the confusion, because the word sounds like it means "caused by genes" and means something narrower and stranger. Heritability is the proportion of the variation within a population that can be attributed to genetic variation. It is a property of a population in an environment, not of a trait, and certainly not of an individual.

The consequences of that definition are counterintuitive enough to be worth stating separately. A highly heritable trait can be enormously responsive to environment: height is among the most heritable traits measured, and average height in many countries rose by more than ten centimetres in a century through nutrition. If you equalised everyone's environment perfectly, heritability would rise toward one, not because genes had become more powerful but because the environmental variation had been removed from the denominator. And a heritability estimate from one population says nothing reliable about another, which is where most of the abuse of these numbers occurs.

Epigenetics is the other term that arrives pre-distorted. The uncontroversial core is essential: chemical marks on DNA and on the proteins it wraps around determine which genes are expressed, and this is how a liver cell and a neuron, carrying identical DNA, become and remain different cells. Development is largely a story about these marks, and cancer is partly a story about them going wrong.

The contested part is transgenerational inheritance in mammals: the claim that an ancestor's experience leaves marks that are passed down and affect descendants. Between generations, most marks are stripped in two waves of reprogramming, which is what makes it possible for a fertilised egg to start over. Some evidence of escape exists, mostly in plants and worms, some suggestive findings in mammals, and a popular literature that has run far ahead of all of it. Your grandparents' hardships are not straightforwardly written into your genome, and the version of the claim that circulates is considerably stronger than the version the evidence supports.

Which leaves a practical rule for reading any genetic finding. A study reports that a variant explains two per cent of the variance in educational attainment. That is a statement about the spread of a distribution across a large sample, not a mechanism operating inside a person; the distributions for carriers and non-carriers overlap almost entirely, and knowing someone's genotype tells you next to nothing about them. It may not even be about the person's own biology: a variant associated with a parent's educational attainment can act on a child through the home rather than through the child, which studies comparing siblings have repeatedly shown. Variance explained in a population, and causes operating in an individual, are different things, and the distance between them is where most of the misreporting happens.