Understanding the Adaptive Landscape
Вставка
- Опубліковано 18 вер 2024
- In 1932, Sewall Wright introduced the concept of the adaptive landscape, and it has been an enduring metaphor in evolutionary biology ever since. In this video, I explain how the landscape metaphor helps us to conceptually understand the interactions between selection, drift, and mutation, as well as sheds light on concepts like pleiotropy, the genetic load, and the evolution of the distribution of mutation effects, as well as bridges the gap between micro- and macroevolution.
#evolution #education
To me, the most mind bending thing in the concept of a fitness landscape is the fitness landscapes of DNA sequence space.
Every "coordinate" in this space can only have 4 different values (ATGC), and there are as many dimensions/directions as there are basepairs in the total length.
That means that the shape of this landscape for a sequence of length L is L-dimensional.
In the graphical representations, these landscapes are always 2D, with many places to go to.
A piece of DNA of only 1000 bases long, would have 1000 spatial dimensions (plus one for fitness), but only 4 places to be in each of those 1000 directions.
A weird quirk about this, is that the maximal possible distance between any two genotypes is equal to the length (so it's linear). Changing thousand A's to thousand C's takes thousand mutational steps.
While the amount of possible genotypes scales with 4 to the Lth power, which is sickeningly fast.
In a two dimensional plain of trillions of square meters, the distance between two random points on average will be enormous. But in a 1000 dimensional space with 4^1000 possible locations, any two random locations are about 750 steps apart.
Things like this warp our intuitions about how fast things can evolve or new functions be found. We can think in 4-D, not 3.2 billion-D.
The combinatorial nature of the landscape indeed make it staggeringly vast. In empirical studies, we simplify it some by considering codons instead of basepairs, which reduces at least protein space by a third. And instead of a topographic plane, the really complex ones are often shown as networks, which has the nice quality of being capable of portraying "resistance" (or mutational bias) as a function of the length of the connections between nodes. But yeah, they're basically unreadable and just make for pretty figures. I like the landscape metaphor because it's one of those simplifications that can elucidate relationships, but if it were represented in a "real" way, it'd be incomprehensible.
I really wish we'd gone over this in my undergrad because it's so deeply fascinating and it ties so many ideas together!
45:45 I'm reminded of an old joke that ends, "I don't need to out run the tiger, I just need to out run you"
Hey Zach, thanks for this video. Really informative and accessible. Definitely keep up with the super in depth mathematical ones, but these are a huge help to small minds like mine.
Also, this is the type of video I would like to share with my young earth creationist friends to help them understand the concepts. Your more technical ones are great for “here is something to show just how much more rigor there is than you ever knew”
Thank you for this amazing explanation. I have just subscribed to what I think is my new favorite channel.
It's interesting (and understandable) how fitness landscapes have fitness peaks to demonstrate local maximum fitness, but I also think that higher fitness could be illustrated as a greater magnitude negative quantity, so thathigh fitness could be seen as a well or trap, like gravity wells. It doesn't change what it illustrates, but it is a metaphor physicists would like. Maybe this isn't an original thought, but I just image a marble drifting in flat, low fitness space until it gets close to a well.
The one issue I see with the "well" instead of "peak" analogy is a well makes it seem inevitable that if the population mean reaches it, "gravity" (=selection) will do all the rest of the work and nothing can stop it from falling in. But selection has alternate forces (drift and mutation bias) that are tugging at it, so climbing up a hill (I think) captures this struggle between opposing forces. If selection is not strong enough, for example, drift will drag you off the hill. But it's hard to visualize that process if the best place to be is in a well, since the imagery seems to indicate once you're there, it'd be a lot of work to move you. The opposite is really the case - when you have the optimal phenotype, all mutations act to move you away from it, and since mutations are constant selection has to work hard to keep you optimized.
I’ll have to refresh my memory and reread Climbing Mount Improbable where I first heard this concept. Mind you I read that book 2 decades ago….. this will be an excellent refresher and probably update my knowledge on this! 🥳👍
Have you thought of doing a video on Dembski’s No Free Lunch ideas? I would love to see that!
Hey..i learn so much from your videos...Thankyou..can you also make a video on evolutionary quantitative genetics explaining the concepts of heritibility, breeding values...i have always struggled with understanding it...
45:25 I just noticed when you said 'difference' you made the ASL sign for DIFFERENT/BUT.
Hey Zach, can you help me in giving an intuitive understanding on how and why does dominance and epistasis make the fitness landscape rugged?
Sure, to understand this start with thinking about why a trait with a purely additive genetic basis produces a smooth landscape. For a two-locus example, A and B, you could imagine aabb is the least fit, Aabb is slightly more fit, AAbb even more, AABb a bit better, and AABB the best. This is a smooth increase in fitness with each additional allele, and selection can "see" the fitness differences between each.
Now imagine the case of dominance. If a and b are dominant, then AaBb is not superior to aabb, despite having both superior alleles. This makes the landscape rugged, because the paths for selection to detect the superior alleles are not straight - it can only see genotypes with double A or B. Epistasis does something similar, but it is the interaction between the alleles that matter. Perhaps A is superior to a *except* when it's on the same genomic background as B, in which case it is worse and a is better.
Hope this helps!
Hi Zack, thank you so much for your work and your content!
I was discussing with a creationist biologist about fitness landscapes and he stated that fitness isn't even used as a variable in biology any more because fitness is supposedly impossible to define and to measure.
Is there any truth to his claim, or was he pulling it from his butt?
His butt. We measure fitness all the time. While there are debates about which components of fitness are most important, there is general agreement about what we mean when we say "fitness".
@@talkpopgen Exactly as I suspected 😅
Most enjoyable.