Are there 'genes for' traits?

It's hard to avoid press stories about 'the gene for' something, because they appear with monotonous regularity. On the day I wrote this, the media reported the discovery of a 'gene for longevity'. The story clearly implied that lifespan is pretty much fixed at conception - contradicting stories elsewhere in the paper about how lifestyle affects health and longevity.

But are there not 'genes for' some characteristics? It depends, of course, what one means by a 'gene for' something. Richard Dawkins, who wittingly or otherwise has done as much as anyone to give people exaggerated views of the power of genes, is careful how he defines this expression. A gene for X, according to Dawkins (1), is any bit of DNA that makes an organism possessing it more likely to have trait X than an organism with some other DNA sequence.

This definition could be useful in some technical contexts, particularly in population genetics models. But, exported into popular discussion, it is almost certain to mislead. Two reasons for this: first, using Dawkins' definition, there are no genes for traits universally possessed by a species. It refers to the likelihood of possessing a trait, and the probability of having a universal characteristic cannot be increased. So, using this definition, genes are not responsible for building organisms - they are merely the explanations for differences between them.

Second, almost every trait that varies between organisms is affected by many genes. Endless confusion results when 'gene for' is used to mean any of the many genes that affect one trait, or is taken as a sufficient explanation of the differences between individuals in a species, let alone as an agent causally responsible for producing the trait.

Further confusion in popular discussion occurs over the tricky, but important, case of genetic influence on continuously varying traits like height or weight. There are probably many genetic and other factors (e.g. diet or exercise) that contribute to how fat or thin people are. This is well-known to plant and animal breeders who try to identify quantitative trait loci (qtls) - genetic markers that indicate genetic variants affecting the quantitatively varying traits they wish to modify. But qtls can only identify genetic variations with statistical relevance to weight because we know the maturation speed of wheat or human weight is affected by many different causes. They can't tell us if the trait is 'genetically determined'. So, if interpreted as a 'gene for obesity', and hence the cause of obesity, this can do nothing but confuse.

With so many factors affecting these traits, a gene that satisfies Dawkins' definition of a 'gene for' weight may have a marginal role in controlling a person's fatness or thinness. And it may be involved, perhaps more centrally, in producing many other traits. So, outside of quite technical contexts, knowledge of this gene does little or nothing for our understanding of weight. The processes through which traits like weight develop are too multi-causal and distributed for anything to be a privileged central cause.

Bearing this in mind, if we turn again to Dawkins' definition, we can see why the term a 'gene for X' is unhelpful. If we suppose that a 'gene for' X must be a central and, perhaps sufficient, cause for trait X, there can be no 'genes for' most complex traits.

But surely there are lots of cases where we do know of specific genes, often definitively located, that determine particular phenotypes? Yes, but these traits are exceptions to the rule in some way. Some are sufficiently superficial that a single different protein may produce a distinguishably different phenotype, perhaps coat or eye colour. But mainly–and these are interesting and important cases–what we refer to as genes that determine particular phenotypes are actually errors in genes. They are alterations that prevent the production of a particular functional protein, generally with a cascade of pathological consequences. Such are the genetic conditions that produce cystic fibrosis and so on, appropriately named single gene disorders. It is worth noting that these errors do not have a specific genetic signature, but comprise any change to the standard sequence that prevents its normal function. Over a thousand such changes are known in the case of cystic fibrosis.

Just because defects in certain genes generate specific disorders in the body does not mean that the same genes control the body's healthy functioning. To take a non-biological example, if you cut the fuel line to a car, it won't start. But it would be wrong to infer that the function of the fuel line was to make the car start.

An example of this fallacy was something I read many years ago about the genetic causes of psychological conditions. This appealed to the case of PKU (phenylketonuria) - a widely used example of how genetic causes interact with external factors, because a restricted diet will greatly alleviate the bad consequences of PKU. Without these precautions, PKU will produce characteristic mental retardation as toxic concentrations of phenylalanine accumulate in the brain. This demonstrates that genetic defects can damage brain development, but it would be wrong to infer that the same genes affected in PKU determine healthy mental development.

In summary, the term 'gene for trait X' is diverse in meaning and potentially highly misleading. The cases where it really does determine X are of very specific kinds, and there are many familiar contexts to which this assumption cannot appropriately be extrapolated. The uses that appear most often in the media, though they can be legitimised by technical definitions such as Dawkins', have no causal implications of the kind typically assumed by the casual reader. A great deal of confusion could be avoided if the expression were avoided altogether.