Comment by jakear | Hacker News Reader

Comment by jakear | Hacker News Reader

I’ve become increasingly interested in what I think would be called “computational botany” if it were to be fleshed out. In looking at this and articles like it, and speaking with both trained and amateur botanists, a common thread is that the appearance of a plant has become increasingly detached from its scientific classification. A frequent refrain in nature hikes I’ve been a part of goes along the lines of: “this plant used to be family X, but then dna testing showed it was actually family Z” - “what? You’re kidding, that’s family Z? It has all the characteristics of an X!” - “Yep, Z. I don’t much understand it myself but apparently that’s what the dna says.”

I think this is a bit silly. It’s akin to categorizing programs by the libraries they use or the programming language they’re written in rather than the problems they solve. What’s more useful to the layperson: “VSCode, IntelliJ, and Vim are Text Editor family; Slack, Talk2Me, and mIrc are Chat family” or “VSCode and Slack are Electron family; IntelliJ and Talk2Me are Java family; and Vim and mIrc are C family”?

The idea with my proposed “Computational Botany” is that the only information used to classify a plant is that same information that would be used to generate a computer model of the plant. So if a specimen has all the markers of an X appearance wise, such that implementing its clone would take the least changes when starting from an existing X, it’s an X.

Of course, the actual implementation of these models may end up requiring an unattainable God-level intelligence, but I leave that as an exercise to future me (or you!)

The phylogenetic tree of life is organized not on the characteristics, but on the evolutionary origin of species. Species are in a specific branch of that tree if they share a common ancestor species with the whole branch. Shared characteristica were a good proxy for ancestry back when the mechanism behind inheritance of genes was not discovered yet. Of course, the concept of a species is a bit fuzzy (even the possibility of genetic compatibility for reproduction as a criteria has issues), but over long timescales the relationships should align into a tree.

A big underlying assumption of this is that genetic material is only passed from progenitors to offspring. Specifically, horizontal gene transfer (common for bacteria, viruses and possibly others, but believed to be very rare for multicelled organisms) throws a huge spanner into efforts to derive phylogenetic trees. It can still be useful though under the assumption that not all genes are eligible for being transferred horizontally.

Also organizing programming languages into a phylogenetic tree is doomed to failure because in general it's quite easy to transfer features from one language to another, and back, as languages are not picky about incorporating features from offspring languages.

I understand how and why they have the naming system they do, I just am claiming that it’s become less and less helpful for amateur identification in favor of what scientists consider useful. I propose an alternative naming convention based only on the expressed characteristics as discoverable by the human eye rather than whatever DNA happens to be in the plant and how that may or may not relate to whatever DNA happens to be in other plants. Basically just demoting the value of DNA in favor of things you can actually see. This clearly wouldn’t replace the existing classifications, merely provide an alternative for folks looking for more explainable names.

Just like how a computer scientist might classify computer programs by instruction set, frameworks, libraries, etc. (things that show up clearly in the machine code but cannot be easily observed from the surface); but a layperson would just use UI and available features (things that are nearly indecipherable from the machine code but immediately obvious on cursory glance)

> The phylogenetic tree of life is organized not on the characteristics, but on the evolutionary origin of species.

My mother's old university biology textbook presents this as one of a few "equally valid" visions for how taxonomy should be organized.

I agree with the modern view that it's the only valid vision, but it's interesting to see it presented in a "teach the controversy" way, and we can see from the comments above that other views have their defenders today.

However obvious gross physical characteristics may appear, evolutionary closeness is much more informative as to most questions. It tells you when things tend to be the same by default, as opposed to being the same due to similar environmental constraints.

Think of evolution as an optimization algorithm that finds organisms for ecological niches.

The monophyletic grouping approach tells you something about the path through solution space taken by the algorithm.

The paraphyletic approach instead focuses on properties of the solutions.

I think it's fine to think of those as equally valid.

> I think it's fine to think of those as equally valid.

Only for extremely limited purposes. Both approaches are about properties of the organism as it currently exists. The paraphyletic approach will let you know about similarity in maybe three dimensions of a ten-thousand-dimensional space. The monophyletic approach captures the other 9,997.

> Both approaches are about properties of the organism as it currently exists.

Not really, no. The monophyletic approach does that only indirectly.

Think of marsupial/mammal convergence. Is it really wrong to think of marsupial mice and mammalian mice as a somehow similar kind of thing?

Or consider secondary unicellularity. Losing multicellularity is as drastic a change as can be, and it makes no sense to try to understand the ecology of a unicellular organism by grouping it with its multicellular ancestors.

Living things are not the way they are just because of their evolutionary history, but also because of the ecological niche that evolution optimized them for.

When similar niches are filled by organisms that are not closely related, that often leads to similar solutions, which is interesting in its own right.

> Think of marsupial/mammal convergence. Is it really wrong to think of marsupial mice and mammalian mice as a somehow similar kind of thing?

It seems to me that most of the similarity can be explained by their shared phylogenetic history. Before the K-T extinction, that ecological niche was probably occupied by a very different small animal, and its occupation of that niche may have even been quite stable. So, occupation of a particular niche by an organism doesn't really imply optimization over the space of possible organisms, but rather over the space of relatively cheap modifications to existing body plans.

Think of it this way - say Australia had never been connected to the rest of the continents, so no organism with a common ancestor with mammals could exist there. How similar would the organisms filling the niche of marsupial mice be then? Probably some strategic similarities like burrowing etc, but probably enough to pass as a pseudo-mammal.

> Think of it this way - say Australia had never been connected to the rest of the continents, so no organism with a common ancestor with mammals could exist there.

This was more or less true of New Zealand until the 13th century. The only indigenous mammals were bats.

> Think of marsupial/mammal convergence.

You are aware that marsupials are mammals?

Not for the same purposes: you are comparing the biological equivalents of history and engineering.

If your primary concern is that of the physical characteristics of the plant, then you don't need to use the classification system based on the phylogenetic tree. You can choose different classification systems based on need!

Any suggestions?

I don't know what your needs are, so I can't help you finding a classification system.

Weird use of the term "computational". More like "visual", maybe.

True, though I’ve explained it a bit backwards here. The true goal is to find algorithms that produce meshes which visually mimic plants. The thinking above is that doing so would result in a new visual classification phylogeny. Such a classification is not the goal, just a happy result.

Reply on news.ycombinator.com