These Sweet 3D Fitness Landscapes Show Evolution At Work

Fitness landscapes are an effective way of showing how well organisms are doing at reproducing. Traditionally, however, these models are static snapshots of processes that often take millennia to unfold. Now, computational biologists have created a video that visualizes these adaptive landscapes over time.


Evolutionary biologist Sewall Wright came up with the idea of fitness landscapes back in 1932. His maps displayed reproductive success, or fitness, of individual organisms as a function of genotype (a organism's genetic constitution) or phenotype (or organism's observable physical characteristics). Traditionally, these maps are shown in two-dimensions, looking something a bit like this:

Illustration for article titled These Sweet 3D Fitness Landscapes Show Evolution At Work

Or this:

Illustration for article titled These Sweet 3D Fitness Landscapes Show Evolution At Work

So, an organism with "higher" fitness has a "higher" chance of reproducing, hence the topographical nature of the maps. Populations, therefore, tend to evolve towards higher ground in the fitness landscape (which isn't entirely intuitive given that topographies tend to show more difficulty attaining higher ground, but there you have it).

Now, thanks to a set of new computer visualizations from Bjørn Østman and Randy Olson, we can see these landscapes in three-dimensions over time for a more illustrative perspective.

Here are some gifs that were provided to io9 by the authors:


Their new models explore three phenomena in evolutionary dynamics that can be difficult to comprehend. As they write at BEACON:

First we show dynamic landscapes with two fluctuating peaks in which the population track the peaks as they appear at difference locations in phenotype space. We also demonstrate negative density-dependent selection, which causes the population to split into distinct subpopulations located on separate peaks, illustrating how speciation can occur in sympatry. Lastly, we show the survival of the flattest where the population prefers a tall narrow peak at low mutation rate, but moves to the lower but wider plateau at high mutation rate. These examples highlight how visualizing evolution on fitness landscapes fosters an intuitive understanding of how populations evolve.


[ Via BEACON via Motherboard ]



I don't think those visualisations work very well. For starters, the population seemed to remain a constant size. So, they have a certain size population sitting on top of a red peak, indicating a species that is successful in a particular niche, well adapted to the area. Then the peak disappears (ie, the environment changes) and that population finds itself in blue territory - that is, not very fit for the environment. Most of them would die off, and only a few lucky ones would make it onto another peak, where they would reproduce.

I think they also missed a good point about stable environments, variety in a populations genetics and chane (although maybe it's because the video was short). If you have a peak with a small population on it, they're centred near the top, huddled together. If the peak disappears, it's most likely the species will too. If the species has hung around for a while, and has a lot of individuals and gene variety, and the environment is accommodating, they'll be spread down the sides of the peak. Those near the bottom will be less fit than those at the top, but they'll still exist and reproduce. If the peak disappears, it's those organisms on the edge which are most likely to make it to a new peak.

Finally, the speed of the changing peaks is important. If a peak suddenly disappears and another one appears quite a distance away, the species probably won't be able to adapt. If the peak disappears slowly, and there are temporary intermediate peaks, the species will be able to adapt. Environments change all the time, it's the speed of the change that heralds catastrophe.

All in all though, I like the graphs. I tried making a similar point with a pencil and paper and didn't get very far...