Could selection link evolution more closely to physics? – Ars Technica

Usually, when someone starts talking about the interaction between evolution and physics, it’s a prelude to a terrible argument that tries to claim that evolution can’t happen. So biologists tend to be a little wary of even serious attempts to theorize about bridging the two fields.

However, October this year saw two papers claiming to describe how one of the key elements of evolutionary theory – selection – fits into other areas of physics. Both papers were published in prestigious journals (Nature and PNAS), so they could not be summarily rejected. But both are very limited in ways that are perhaps a product of the interests and biases of their authors. One of them may be the worst-written paper I have ever seen in a major journal.

So fasten your seat belts, and let’s dive into the world of theoretical biology.

More assembly required

We can start with the terribly written paper. It introduces assemblage theory, a potentially useful way of thinking about what natural conditions can help Combinatorial chemistryThis results in the generation of a complex mixture of elaborate molecules. But this is not how the authors, many of whom are chemists, present the idea.

The first sentence of that paper makes it difficult to make evolution consistent with physics: “Scientists have grappled to reconcile biological evolution with the immutable laws of the universe determined by physics.” this is not true. Evolution is perfectly compatible with physics, and we’ve known that for a long time. It is impressively untrue The paper they cite In support of it, he only mentions physics once, just to say that people have some misconceptions about it.

Things don’t get better from there. “These laws [of physics] “They support the origin and evolution of life and the evolution of human culture and technology, yet they do not anticipate the emergence of these phenomena.” This is true in the sense that any emerging phenomenon is, by definition, difficult to predict on the basis of the behavior of its simpler components, but this does not mean that we need a new theory to relate it to fundamental physics.

However, we get one. “We present assembly theory as a framework that does not change the laws of physics, but rather redefines the concept of the ‘object’ on which these laws operate,” the authors claim. But they never did, despite a section of the paper titled “Clustering unifies selection with physics.” The targets of assembly theory can be things such as atoms, which can be easily analyzed using the laws of physics. But they can also be intangible things like concepts, with researchers reporting that human languages ​​and memes are likely amenable to clustering analysis.

Therefore, the whole “unifying choice with physics” is, at best, a distraction and effectively interferes with the explanation of assembly theory. Accordingly, the paper does a terrible job of explaining this. However, somewhat surprisingly, the theory can be easily explained in a series of less than two dozen social media posts, Proved by Carl Bergstrom.

Neglecting physics

As Bergstrom notes, aggregation theory works best if you think of it in terms of chemistry. Suppose you mixed a mixture of simple chemicals together and let them react. The likely result is a mixture of polymers, each assembled from a random mixture of simple chemicals. You’ll have a lot of molecules, and each one of them will be distinct. But what if that’s not what you see? Alternatively, you may see that a limited number of groups were highly favoured. There will still be a lot of molecules, but they will all match one of a few templates.

This is the situation we see with proteins. With 20 amino acids that can combine with each other in any order, even a group of proteins 50 amino acids long can contain a large number of individual molecules. But the truth is that we only see a small portion of this potential pool, because evolution has selected a limited number of functional proteins.

Assembly theory posits that any set of molecules can be viewed as a combination of the minimum steps needed to assemble it—the history of how it got there—and the number of copies present. The higher this value, the stronger the specification required to produce it.

As with actual evolutionary processes, this recognizes that the final population is a product of the history and contingencies involved in the first steps of assembly. It is potentially useful in two ways. It provides a method for quantitatively distinguishing between mixtures of randomly assembled polymers of different monomers, polymers that are the product of ligation of many copies of a single monomer, and polymers produced by selection. As long as the steps and number of copies can be measured, the amount of selection that has been involved in producing a set of molecules can be measured.

What it doesn’t do is unify any of this with physics. The authors acknowledge that in the body of the paper, they write, “combinatorial spaces do not play a prominent role in current physics, because their objects are modeled as point particles rather than combinatorial objects.” “This definition is, to some extent, inconsistent with standard physics, which treats objects of interest as fundamental and unbreakable.” But none of that stopped them from writing exactly the opposite in the summary.

Lay down the law

The second paper was written by a team that includes a group of astronomers, and it shows. Its focus is on finding parallels between selection in evolution and other processes that build complexity. The examples she uses are things like the construction of increasingly complex mixtures of elements in stars and the increasing complexity of the minerals that form in planets – things that are of great interest to astronomers and planetary scientists.

Part of the paper involves identifying similarities between these systems. “Sophisticated systems appear to be conceptually equivalent in that they display three salient features: 1) they are composed of many components that have the capacity to collectively adopt vast numbers of different configurations; 2) there are processes that generate many different configurations; and 3) configurations are preferentially selected on the basis of function,” the authors write. In general, the evolution of all these systems is also driven by energy dissipation.

Details may vary, but the authors argue that the similarities suggest that natural law is appropriate for describing behavior. The law they arrived at is:

Systems of many interacting agents exhibit increased diversity, distribution, and/or patterned behavior when multiple configurations of the system are subject to selective pressure.

But, of course, there are a lot of things that are not parallel. Evolution is constantly exploring new configurations, while the formation of elements and minerals is limited by physics and chemistry, respectively. While these systems can explore different pressure and temperature regimes, they are very limited compared to biology. As the researchers acknowledge, “Recent work has estimated that the fusion phase area of ​​Earth’s current biosphere greatly exceeds the fusion phase area of ​​the abiotic universe.”

Although it has never been well defined, the authors acknowledge that biology appears to contain what is called “functional information.” In other words, when something “works,” biology has the ability to continue producing that thing and generating variants of it. While this is somewhat similar to the equivalent of an atomic nucleus or a stable metal, it lacks the external storage of information that DNA provides.

Overall, the article is much better written, and its arguments, which are a bit more limited in scope than those of assemblage theory, are easier to follow. But at the same time, it is unclear whether these arguments really support the case that the similarity between these examples is deeper than conceptual similarity.

Is any of this helpful?

Natural laws tend to be both conceptually and empirically useful. Things like Newton’s laws of motion and Mendel’s laws helped organize thinking in a way that led to many useful experiments, and those experiments eventually led to identifying cases where the laws broke down. This led to further advances such as relativity and genetics.

However, this type of process has been going on for centuries, so it is difficult to judge whether the proposed natural law can achieve something similar. There are no ways to use this to push beta software that I can see at the moment, but that doesn’t mean someone won’t develop software eventually.

When assembly theory is taken at face value—a way of looking at complex chemistry and its consequences—it has the potential to be useful, as there are clear ways to pursue it experimentally. Since building complex molecules was a fundamental part of the origin of life, there are important questions that can be applied to it.

But it also aims to apply to any selection process — not just the evolution of organisms, but also “cell morphology, graphs, images, computer programs, human languages, memes, as well as many other things,” according to the people proposing it.

For some of these items, it is possible to understand the history enough to know the assembly process or estimate the minimum number of steps required for assembly. But it’s unclear whether this is possible for things like evolution or some other aspect of the origin of life. No one knows exactly what features of life were part of the first cells or how they were assembled, so it’s not clear that assembly theory can be used there. The same is true for the full set of genes found in the common ancestor of all life forms today. Differences between related species, such as mammals, appear to be largely the product of mutations that lead to subtle changes in gene activity, which are difficult to identify and characterize.

None of this is to say that clustering theory is wrong, but the challenge of obtaining the information needed to put it to use may range from impractical to impossible for many important questions. Figuring out how to use them effectively in situations beyond chemistry will be a real challenge. Unfortunately, the people proposing it claim that it addresses problems that do not exist and cannot be addressed by it, so I expect the challenge will be much more difficult than it needs to be.

Nature, 2023. DOI: 10.1038/s41586-023-06600-9
PNAS, 2023. DOI: 10.1073/pnas.2310223120 (About digital IDs).