They say if it talks like a duck and walks like a duck, then you have to conclude it's a duck. That's a snide way of saying that you judge things by how they look, not necessarily what someone says--and is often applied to politicians' obscurantist rhetoric, for example. But it can have implications for biology, too.
If it has a beak like a duck, it's a duck....unless it's a genetically modified chicken. Experiments with two genes, Bmp4 and Cam1 (sometimes written CaM), have shown that not only are they expressed in critical areas of the embryonic jaw, but that altering their expression sites, timing, or intensity can alter jaw (and hence beak) size and shape. Famously, this is an explanation of variation in the Galapagos finches, but the involvement of the genes in many species, including mammals, is clear. An article in Annual Review of Genetics (Roles for BMP4 and CAM1 in Shaping the Jaw: Evo-Devo and Beyond, J. Parsons and R. Craig Albertson, vol. 43, Dec 2009) summarizes the extensive knowledge and experimental results related to these two genes and jaw development.
So, experimentally altering the genes has effects on face and jaw length and shape, as can natural alterations. To some extent, at least, give us the alleles at these loci and we can predict the face--especially if the alleles have major effect. Of course they're not the only contributing genes, so this is a question of penetrance. That's the probability of a jaw effect given the genotype.
But what about the inverse question, known as the detectance? If you give us the face, can we identify the contributing genes? This is a very important question because natural selection sees faces, not genes. It could be of medical importance, too, if the presence of a disease points to its specific cause in a way that may help choice of therapy.
If many genes contribute, as we know they do, their diverse effects may be difficult to assess and a given trait, like a long vs short face, may not need to involve any particular subset of these genes.
A very fine recent paper by Roseman, Kenny-Hunt, and Cheverud (Phenotypic Integration Without Modularity: Testing Hypotheses About the Distribution of Pleiotropic Quantitative Trait Loci in a Continuous Space, Evol. Biol. 36, 282-191, 2009) provides an interesting case in point (the first and last authors are people with whom we have an active research collaboration, and hence we're predisposed in their favor, but the paper is fine on its own merits!). They crossed two standard strains of inbred mouse, called Large (Lg) and Small (Sm), then intercrossed the offspring for 10 generations (producing the'F10' offspring). Then, they landmarked 15 different positions on the jaws of 1,240 F10 mice, and computed inter-landmark distances. Each mouse was been genotyped for 1,470 variable sites (SNPs), or 'markers', spaced more or less evenly across all the chromosomes.
Roseman et al. did the GWAS (genomewide association study)-like mapping experiment: asking for each measured distance, in what parts of the genome marker variation was statistically associated with variation in the distance. They found 28 such chromosomal 'hits', each associated with one or more distance.
What's relevant to this posting is that neither Bmp4 nor Cam1 (called Calm1 in mice) was located in any one of these candidate chromosome regions. Now statistical data have many problems and limitations, and more data may lead to revisions (which will be forthcoming, since F34 generation mice are available, and will provide much more refined map locations).
But the point here is that while nobody doubts that Bmp4 and Cam1 are involved in mouse jaw development (expression studies and other work makes that manifestly unambiguously clear), the genes seem not to be materially involved in the jaw size or shape differences between these particular mouse strains.
The reason is very simple and can be characterized as phenogenetic (or genotypic) equivalence. A fundamental characteristic of complex traits is that many different genotypes can produce essentially the same phenotype. Even if a genotype were accurately to predict phenotype (which we know is only true a small fraction of the time), a phenotype does not as a rule predict the underlying genotype--that's the major lesson of GWAS experience!
In evolutionary biology and systematics, similar traits in related species are called homoplasies if they have different causes--if they evolved independently and, in modern terms, used different genes. They are classically called analogous rather than homologous.
In relation to today's post, this means that similar faces in different species may be homoplasies. Yet in other cases, even among the same species, they may be homologous--due to mutations in the same genes. Or, if the genes are the same but the mutations different, they are a kind of mix.
This can confound evolutionary understanding if one is not careful, because the same differences--different genes same trait variation, or same genes, different alleles, same trait variation--can occur. The reason is that the evolution of variation and of speciation is a process, not an event.
Predictance and detectance are very different entities that can, but need not, have similar values. They are fundamentally important in the current debates about gene mapping in biomedical genetics, and in evolutionary genetics. We'll have more to say in other posts, as this is a basic and important subject!