Gene Models

Beau, aka "snow pig", after a hard day of plowing show with his snout.

I’m in Virginia this weekend where I’m currently watching snow accumulate and my little snow pig root around.  The ruler is reading around six inches, which in central Virginia is enough to lock the grids and cause chaos. Guess I’ll have to lounge on the couch all day.  Shucks.
 
Readers of my post from last week may recall mention of some genetics models my computer was slowly cranking out. Yep- those are still running.  But it’s close to finishing and I think I can predict what the results are going to be. It’s potentially really interesting, but, I don’t want to report any findings, no matter how preliminary, until I’m confident they can hold at least a little weight. Plus, I’m meeting with some collaborators next week and I think they will recommend a few tweaks to the analysis that will help clarify some oddities and rule out other possibilities.  But, soon.  Promise.  
 
In the meantime, I’ve started the early stages of preparing the publication that will formally report the results from this study. This process is always a little daunting. It usually starts with a blank Word document, a lot of distractions and procrastination, and many frustrations as I try to find the perfect words and the perfect topics to tell the perfect story for the data.  I call this stage “unproductive circling,” and for me it can last for days. Then, I break down and just start typing. The resulting text is usually a horrible mess of incomplete thoughts, poor grammar, cuss words, and confusion. But, it stops my wheels from spinning so I can get traction and move forward. From there it’s all about refinement- reread, rewrite, reword.   Over and over until the paper you were so excited about becomes utterly boring and dull. That’s when you know you’ve done your job correctly. 

---

In case you're curious, this is an example output from the models that are running on my computer. I'll explain what these mean later, but for now, populations (numbered at the bottom) are more genetically distinct if they are shown as a solid block of color. So, for example population #3 is more genetically distinct than population #4, which COULD indicate that #3 is more isolated than #4.

For those unfamiliar with scientific publications, it starts with an introduction.  Usually, it’s a couple pages in length explaining what the scientific community already knows and justifying why your research project was needed to plug an information gap. Many writers will disagree, but I really enjoy writing introductions.  There are millions of research articles out there, and it’s your job to figure out why your lone study is still important to science.  It’s a messy thought exercise that ends in a succinct story.  It’s as close as scientific writing gets to poetry.
 
Much of this week was spent “circling.” I did all I could do to avoid writing, but I finally opened the blank Word document and stared. The problem I had was trying to decide what about my project was interesting.  I know that sounds a little backwards. After all, shouldn’t you know your research is interesting before you do it?  Yes, but that’s often not how research works. It’s only after it’s all done that you realize you got an unexpected result, or you collect data knowing it will be of interest to someone, but don’t know who that someone is until you’re nearly finished. So, I kept circling between two fundamental questions- is this study interesting because it’s brook trout, or is it interesting because it’s a genetics study of an aquatic organism?
 
While the answer is both, it turns out that the impact of the study is probably further reaching if I stretch my mind beyond brook trout.  Sure, the study will be important for brook trout management. But, can it tell us more about genetics in general?
 
As I struggled with this question I do what I normally do- I started chatting with someone online. Luckily for me, my friend, Will, happened to be on. He studies genetics in many several wildlife species and knows far more about the topic that I ever hope to. He has been my go-to person during this process because he can give me insights into why something isn’t working or a next step. As we were talking he turns my attention back to a paper that has been sitting in my “to read” library for a long time.
 
To be honest, there is nothing earth shattering about the paper. It describes ways organisms (and their genes) can distribute throughout streams. But, the paper is incredibly necessary because streams are a special type of habitat. Many genetics studies are conducted on terrestrial organisms that can disperse in any direction.  But, aquatic organisms are restricted to streams (for at least part of their life) and so have very different controls on how, when, and where they can move.
 
In some ways, you’d think streams make genetics models easier.  Movement can only be up and down. But, that’s not entirely true for organisms like salamanders and macroinvertebrates that can fly or walk on land. Even for fish, which really are restricted to in-stream movements, models easily get confusing when you meet a tributary. Now movement isn’t just up and down, but the fish can pick any number of tributaries to occupy.
 
Taken together, suffice to say streams quickly complicate gene models. But, this paper was helpful because it packages just about every possible way stream-dwelling organisms move into four models.  (And, don’t get scared away by the term “model.” It this case a model doesn’t involved any complex math or equations, it’s just a possible scenario for how the world works.) 
 
The four models included:

• Widespread gene flow model: Organisms move around the landscape freely, and populations that are reasonably close to one another are genetically similar. It’s important to note that proximity can be measured two ways.  The first is overland distance, or “as the crow flies.” This distance is measured by a straight line between two points. The other measure is stream distance, or “as a fish swims.”  This distance is the shortest distance between two points, but only measured using waterways. Almost always stream distance is greater than overland distance. And ,if you are jumping between watersheds stream distance can get very, very large between two points that are, straight line, very close to one another. In a widespread gene flow model, organisms can readily move outside of streams, and so two populations that are reasonable close (i.e., moderate overland and/or moderate stream distance) are genetically similar. Think about organisms that are good fliers, like aquatic marcroinvertebrates with terrestrial adult stages, and you might have an organism with widespread gene flow.
• Headwater model: Organisms seek headwater habitats, but are able to move overland. In this scenario, organisms in one headwater are seeking the next closest headwater environment. Often time, the closest headwater environment is found in a completely different watershed (i.e., short overland distance, high stream distance). Organisms that conform to the headwater model also tend to be adult stages of flying macroinvertebrates, but headwater specialists tend to have reduced flight capability compared to widespread dispersers.
• Stream hierarchy model: Here, organisms are generally confined to streams and populations within the same watershed tend to be more genetically similar than populations in different watersheds (i.e., high-low overland distance, low stream distance). Fish are the most obvious example of this model, but some species of macroinvertebrates that only move during aquatic juvenile stages and also conform to the stream hierarchy model.
• Death valley model: Here, organisms are trapped in their home habitat and there is no physical connection to other suitable habitats.  In this situation, all populations are genetically distinct even when they are close (i.e., short overland distance, and short stream distance). This model is usually reserved for organisms that occupy isolated pools or areas of spring upwellings in an otherwise terrestrial environment. 

Imagine from Hughes et al. (full link to the paper below). In these images, imagine that there is a stream flowing north, and another flowing south. The dots represent populations where colors represent genetic similarity. Under the death valley model (A), there is no exchange of individuals between populations. For the stream hierarchy (B), organisms move within a waterrshed, but do not cross watershed boundaries (so genetics are similar for populations in the north flowing stream, but different between the north and south flowing stream). In the headwater model, organisms look for the next closest headwater, which is often in an adjacent watershed (so genes are shared among populations that are close in overland distance, regardless of watershed). And, in a widespread gene flow model, distance doesn't prevent movement and all populations are genetically similar.

A beautiful brook trout, your prize for reading through all those genetic models and making it this far in this blog.

Conceptualized models, such as those described above, may not completely explain a single dataset, but it does help quickly put a dataset into perspective. We know that trout shouldn’t conform to the widespread gene flow or headwater models. Both of those require overland dispersal or very, very far stream distance dispersal. 
 
Theoretically, trout should conform to the stream hierarchy model.  We know that trout move, and they are more likely to move within their watershed than outside. And, previous studies have found evidence to support that trout genetics follow other requirements of the of a stream hierarchy model. 
 
But….(and here is why my study is special), most trout genetics studies are done at fairly small spatial scales.  Usually 10 sites that are near one another and that are part of a larger watershed network.  We studied 28 populations located throughout the Loyalsock Creek watershed.  Near as I can tell, there isn’t a published study on brook trout that sampled as many fish from such a large distribution.  So, hopefully, we can tease apart which genetics model most explains brook trout.
 
Why does this matter?  From a practical view, it can help in management.  Management is most often applied at a watershed scale, but that’s not often the scale scientists do genetic studies.  So, our gene data, which is arguably one of the most powerful tools we have, is not informative to our management efforts- at least not without a lot of assumptions. But, if we know that brook trout genes, at a watershed scale, tend to conform to one of the aforementioned models, then suddenly we can make more informative guesses (notice I say guesses) about how trout are moving in watersheds where we haven’t done a robust genetics analysis.
 
From an ecological perspective, understanding which model brook trout conform to could be just the kind of interesting quirk ecologists dream about.  From the telemetry data, we know brook trout move among tributaries.  But, if they still conform to a death valley model, it could mean that they return back to their home tributaries. Brook trout, after all, are closely related to salmon.  Or, it could be a mix of models, in which case we ask why some populations have adopted different behaviors that has then lead to distinctly different models of gene flow.
 
I have a hunch as to what is going on.  But, only time will truly tell…
 
If you’re interested in reading the paper I described above, it can be found here. 

---