Monday, December 23, 2013

Thumbs up...and heads down. More on 'selectionism'

Last week we used my glorious thumb and its several properties to illustrate the difficult problem of defining what a biological trait is, in relation to the widely held assumption that our traits are here strictly because of natural selection.  Our point was that it is quite tricky to define a trait, or specify just how precise, and in what ways, our assumed selection has worked, and is working.

A thumb is seemingly so simple that it can be surprising how difficult this is.  But here is another 'trait' that is clearly an evolved trait with what even I would agree is of fundamental importance to our existence, and in that sense, 'adaptive': the skull.  It, if it is an 'it', clearly has evolved and using evolutionary theory and comparative anatomy, we can identify what we assume are evolutionarily corresponding 'landmarks' (the dots in the figures below).  These are canonical anatomical locations shared by members of a species, but the distance between them varies within and among species.  These measures can be indicative of genetic and/or life experience variation.  But is the skull, or a landmark, the product of natural selection?  To answer that question, we must understand what 'it' is.

Landmarks: human, baboon, and mouse (Figure by Joan Richtsmeier)

This example is courtesy of work with our long-time collaborator (and good friend) Joan Richtsmeier, an expert on imaging and analyzing craniofacial dimensions and their pattern of variation.  Indeed, the figure shows various standard measures that people in this field use.  We and Joan, and Tim Ryan also in Anthropology here, along with Jim Cheverud of Loyola in Chicago, Heather Lawson at Washington University Medical School, Jeff Rogers at Baylor College of Medicine, and others, have been looking at the genetic contributions to craniometric traits in laboratory mice, and in a big baboon genealogy housed in Texas, and to compare that with what is known about human skull-shape variation.

We have CT scanned the skulls and made careful identification of standard 'landmark' points, the dots in the figure (the Figure just shows whole skulls, but in a CT scan the landmarks can be in 3-Dimensional locations and we can look at any slice through the object).  Then, using hundreds of individual skull-scans, we have been relating inter-landmark distances (lines that can be drawn between any two landmarks) to genetic variation, to 'map' the genomic locations that affect each distance's (or correlated distances') variation.  Yes, skeptical though we are of its over-interpretation, we are using GWAS-like methods (QTL mapping) to do this.  (We've described the difficulties with this work a few times on MT, including here and here.  The gene mapping issues we discussed in these posts remain unresolved.)

We chose the landmarks and distances based on what has been done before, or reflecting obvious locations like where sutures join or peaks in curvature and so on.  The statistical methods in the genomic analysis and so on are also well-tested (there are some controversies but they're beside the point, or comparable to the point that we're making here).  For most of the landmarks in one species, homologous locations can be found in the others (in our work, humans, baboons, and mice).  In this sense, the distances are very clearly 'objective,'  replicable measures of shape and size in the skulls.

But in what sense are they 'traits'?  Is the fact that we can identify and measure them, for reasons we decide on, a sufficient basis for the definition of a trait?  Is a measure's genomic 'control', regions that we identify by mapping, actually 'for' that measure?  If so it implies that other aspects of skulls, such as the bone thickness, dental cusp or shape patterns, and so on, are unrelated and irrelevant to the measure.   That seems hardly to be justified.

We know these traits evolved and that what we observe today ancestrally had to be consistent with successful survival and reproduction--passing the screening of natural selection or whatever other types of functional trials were imposed.  And they survived chance events, too.

But just as with the thumb example from Friday's post, it is we the  investigators who decide what a 'trait' is, and if we wish to infer adaptive scenarios it is we who decide what is adapting.  That leaves us the freedom to decide with equal subjectivity what it was adapting for.  Of course we know that other aspects than what we're testing had to have evolved consistently with what we're testing, or else a single, functionally viable and cohering skull would not have developed in the ancestral individuals as they and their various craniofacial 'traits' (whatever they be) were evolving.  But of the countless genes expressed in, and hence clearly contributing to, skull development, how is it that we can decide if the gene is 'for' some particular distance measure?  A gene may contribute to that strictly via some other effect than what we think our measurement is about (indeed, what 'strictly' means is itself questionable).  Indeed, most genes contribute to many different tissues and structures and developmental stages in the head, not just one.

Here, we're not just playing possibly-trivial thumb games, but are considering the nature and evolution of one of our most vital structures.  Yet the same question: what is a 'trait'? arises in regard to essentially every distance we assessed and to the very act of choosing landmarks and distances to measure.   So what we are really considering is the nature of evolution itself.

Chance is a chancy subject because it, and associated 'probability' are very hard to define.  But even if chance has, and always had, nothing to do with our 'traits', and if selection really were, as so often asserted, the be-all and end-all of life (except, of course, for love and lunch), and even if we've had Darwinian ideas for more than 150 years, and even if they were wholly correct, then even then, natural selection and the nature of evolution are still far from completely understood.

Somehow, however, many of people in evolutionary biology don't seem to realize this, or don't want to acknowledge it.  Or don't they care about truth, so long as they can tell a good story?

6 comments:

Kirk M. Maxey said...

It may be helpful to step back and look at traits from a completely different perspective - that of disease trait phenotyping. Sadly, there is no more dismal, unformed and embryonic a science. Let's pick a good, well studied syndrome like Ehlers-Danlos. In lay terms, it means you are double jointed. Mutations in either the actual structural genes of collagen, or in the enzymes that make or remodel it, cause E-D. The observed consequence of E-D mutations are weakened, yet hyperflexible joints, and stretchable, elastic skin that can be pulled away from the underlying tissue. Imagine a person who can reach thelr left arm behind their back, under their right armpit and up to scratch their chin. Or, on a seasonal note, the Bible states that ,"Joseph tied his ass to a tree and walked into Jerusalem." which might have literally been possible if his E-D was pronounced.
To take one of the clinical manifestations of E-D and define it as a trait, one must specify clearly what that is. To take the example of Ken Weiss' thumb, the "trait" would be the deformability of his angle theta under a given pressure. An E-D patient can bend the distal phalanx backwards until the thumb tip almost touches the second knuckle, so theta approaches 180 degrees. At some arbitrary point (90 degrees, let's say) most people with normal collagen would find that exercise excruciating, so we can define the E-D "trait" as anyone who can hyperextend to this point without extreme pain.
Contemplate that the threshold angle, the definition of "extreme pain" and the amount of force applied are completely arbitrary. (I'll tweet you a pic of an E-D patient applying the force with their own index finger.)
At least in the case of this disease, there are perhaps a dozen known genes that have all been implicated in the ability to transmit this trait if properly mutated. If instead one wants to talk about traits that confer good, useful working body parts instead of rubbery, defective ones...where on earth does one start?
I only realized how underdeveloped the science of trait phenotyping was after having been first genotyped myself, and then trying to ask, "So what exactly was it that this genome built?" The known trait descriptions of myself, or any other human, do not extend past a few printed pages, no matter how minute and detailed one tries to get. Blue eyed. 5'10" A+ (Note that only 2 of those 3 are biometric - the last one is biochemical.) Wavy haired? Hyperactive? Optimistic? Again...what exactly is that, and how are you going to measure it?
Until we can accumulate biometric trait descriptions of organisms that are validated and of approximately the same magnitude as our genetic determinants - 3 billion or so - we won't know how to talk about human genetic determinism or the lack of it.

Ken Weiss said...

This is well-put, and thanks.
The issues of correlation vs causation, determinism vs chance, our definition of 'traits' vs nature's and so on are all involved. The idea of reducing causation down to single elements, be they genes or habits, is so appealing that it seems impossible to resist. The reduction pleases both the vendors and the buyers of services, promises, hopes, and so on.

Science is all too human in these regards, and our society too impatient.

If fields like quantum mechanics are any guide, there may not even BE answers to poorly posed questions.

Kirk M Maxey said...

Biometric trait phenotyping didn't end up in its current state of utter disrepute and failure by accident. It has a politically charged and abusive history that must be read to be appreciated. "The Mis-measure of Man" by the late Stephen Jay Gould is a good primer for this topic. But just to give flavor - between 80 and 150 years ago, the main reasons men measured the intracranial volume was to prove that the more evolved Teutonic races had larger brains than the primitive Amerindian and African ones, and to show that men in prestigious professions had bigger brains than those in lowly ones. And let's not forget to also downgrade small-brained women too - it was the style of the day. As measurements and facts accumulated to confound these hypotheses, shoddier measurements and outright data fudging were indulged to maintain the political fictions.

This sorry state of affairs prevailed for a metric that isn't that challenging - measuring the volume of space inside the braincase. We do it with remarkable accuracy today on badly crushed 2 million year-old hominid skulls. I myself have consented to let Harvard MRI my cranium at high resolution, so within 1 or 2 cc this measurement itself is an established fact. But then - to which of my 3,256,067 SNPs, 17 well characterized tests of blood chemistry, 2 known deleterious mutations, one childhood hospitalization, 3 standardized tests of intelligence, and one 3-day suspension from high school do you want to try to associate it?

We've failed in the past here because we jumped too eagerly from simple metrics to unvalidated social and political associations. Thank god the Freudians are all gone and discredited, or they'd be interjecting their random behavioral nonsense into the mix as well. I can just imagine some evopsycho zipperhead aligning my brain measurement in his tabular comparison of intracranial volume and ego dominance.

The challenge for scientists is to compare only real, validated THINGS. Describe them, define them, show that they can be blindly reproduced by others. Here's an example: There is a span of nucleotides on chromosome 17, where the consensus rs62075803 base is C. I myself have two copies of T. It's in an untranslated intragenic bit of dark DNA between the PTRF gene and the ATP6VOA1 gene. But the frequency of this base in the 1000 Genomes is <0.4% - it's pretty rare. Whether my SNP and intracranial volume are associated with one another is a question of database size and computer algorithm - but it's not a question of politics or gender neutralism. This is the sort of place the measurement of man needs to go.

Anne Buchanan said...

Aside from the politics, and of course there are politics, there are very serious epistemological issues here having to do with how we define biometric traits, and they affect searches for genes 'for' these traits, and whether we find them, and what it means when we think we have. It's not just evolutionary questions that are at play, as Ken described, but loads of issues having to do with disease prediction and understanding. We write about these kinds of issues all the time.

If it's not obvious what a measure represents, it's unclear what you're even describing or trying to explain when you look for associated genes. Or, in Ken's examples, evolutionary explanations. Landmarks on the skull are a good example, but so is intracranial volume because, developmentally speaking, it's not clear whether brain size drives skull size, or skull size drives brain size, or both, so you've got a measure, but what does it even mean? Forget the politics, what does it mean biologically? Or genetically?

Is blood pressure a trait? BMI? Or the ultimate in politically charged measures, intelligence/IQ? When there are multiple pathways to outcomes like these, or the pathway is uncertain, or the outcome can be defined in multiple ways, as intelligence -- or autism or heart disease, or schizophrenia… -- we've got problems even before politics enters the fray.

Kirk M Maxey said...

All very good points Anne. After reflecting on this, there are a few general constraints that might be articulated as a general guide for all wishing to publish data and reference the word "trait." But when I tried to list those - I went way past the character limit for your blog. :-)
I do think it would be helpful to start a #humantrait, or similar hash-tag for people to post when they think they have spotted a "thing" about people that implies some genetic component. But also look for something on this soon at kirkmaxey.com.

Anne Buchanan said...

Thanks. Please do let us know when you've posted something. It has been said that if I know where you come down on genetic determinism -- of anything, from how we vote to our risk of dying of lung cancer -- I know your political orientation. I think this is no longer true -- genetic determinism now colors everyone's view of the world, to a greater or lesser degree. And it's largely unjustified. (Here's a piece Ken and I wrote on this for Aeon a while ago.)