Forum:Theoretical Question: Is Sub-Angstrom Geometric Identity (RMSD < 0.1 Å) in Convergent Evolution Possible?

Question

[Alberto]

Dear community,

I am an independent researcher exploring the fundamental principles of molecular motor function, specifically focusing on the classic case of convergent evolution between myosin and kinesin. The long-standing consensus is that their structural similarity is largely topological, not geometrically precise, given their negligible sequence homology. My research, however, has led me to a result that challenges this assumption, and I would appreciate this community's expert insight.

My theoretical question is: What is the accepted lower limit for a meaningful backbone RMSD between two non-homologous, convergently evolved protein domains? Has a structurally identical core, with an RMSD significantly below 0.1 Å, ever been documented?

I ask because my own heuristic-driven, non-homologous superposition analysis has identified a conserved "core-bridge" socket between Kinesin (3kin) and Myosin (1fmv) that yields a remarkable RMSD of 0.096 Å.

This result suggests a level of geometric conservation I previously thought to be impossible, pointing towards what I've termed "Geometric Deep Homology" – a fundamental, immutable design principle that transcends sequence. I am struggling to determine if this is an unprecedented finding or simply an artifact of my methodology. I have uploaded a preprint detailing the full methodology and results to Zenodo to establish priority and invite discussion. The DOI link: https://zenodo.org/records/15970181

Any references, thoughts, or discussion on a finding of this nature would be immensely valuable. Is this a known phenomenon, or have I stumbled upon something genuinely new? Thank you for your time and expertise.

Alberto Renee Mora Peña [e-mail]

[GenoMax]

What is the accepted lower limit for a meaningful backbone RMSD between two non-homologous, convergently evolved protein domains?

While you wait for an answer, in case you have not asked this question to an AI/LLM, the the answer looks to be

For non-homologous, convergently evolved protein domains, a meaningful structural similarity is generally considered when the C-alpha RMSD is:

~2.0 to 3.5 Angstrom over a significant portion of the structure (typically 60–80 or more aligned residues).

At a minimum you will have to include the following (for those who don't want to read your preprint)

my own heuristic-driven, non-homologous superposition analysis has identified a conserved "core-bridge" socket between Kinesin (3kin) and Myosin (1fmv) that yields a remarkable RMSD of 0.096 A.

What is the length of the region you are referring to here.

As an aside, asking GPT to compare the two proteins brings up the following.

Core alpha/beta-sheet + P-loop region: RMSD around 1.5–2.0 A

Full motor-domain alignment (incl. helices & loops): 2–3 A

Despite nearly zero sequence identity, this low backbone RMSD over a substantial structural core strongly supports their homology—pointing to divergence from a shared P-loop NTPase ancestor.

Ref: https://www.cell.com/structure/fulltext/S0969-2126%2896%2900055-X

[Alberto]

GenoMax, Thank you for the excellent and rigorous feedback. That is a critical question. The region I am referring to is a four-residue constellation. The near-perfect identity (RMSD = 0.096 Å) is achieved through a "core-bridge" mechanism (it needs to use backbone super) A rigid Core (2 residues): Val174 (Kinesin) vs. Ile258 (Myosin), which function as identical B-strands. A flexible Bridge (2 residues): Ser175 (Kinesin) vs. Gln259 (Myosin), which connect the core via a loop (Kinesin) and a B-strand (Myosin), respectively, while maintaining geometric integrity.

I appreciate you engaging with this idea

Answer 1

[Mensur Dlakic]

My theoretical question is: What is the accepted lower limit for a meaningful backbone RMSD between two non-homologous, convergently evolved protein domains?

I think you should be asking what is the meaningful domain length that will be accepted, or at least a meaningful number of scattered residues. I don't think 4 residues qualifies on either grounds.

I am not categorically arguing for divergent evolution here as I have not studied these proteins, but I think it is possible. Divergence often leads to proteins that are only 5-10% identical, and at that point it is difficult to prove a common ancestor except at the structural level. We recently had a couple of cases like that:

• https://pubmed.ncbi.nlm.nih.gov/32284351/
• https://pubmed.ncbi.nlm.nih.gov/21441348/

If we consider a number of molecular changes (mutations, indels, rearrangements) that are needed for significant divergence and compare that to a number of changes needed for convergence, it makes sense that we have a million-fold more examples of divergence over convergence. It just may not be obvious how the divergence happened mechanistically. So you may call that "Geometric Deep Homology" but in reality it could be just a divergent evolution with many steps.

Has a structurally identical core, with an RMSD significantly below 0.1 Å, ever been documented?

Maybe I am being nitpicky, but 0.096 is not significantly lower than 0.1. I would need some convincing that 4 residues constitute "a structurally identical core" even if they were catalytic, which doesn't seem to be the case here. I suggest you go through some old papers by Nick Grishin, like back in early 2000s.

https://pubmed.ncbi.nlm.nih.gov/?term=grishin+nv%5Bau%5D&sort=pubdate

[Alberto]

Dear Mensur Dlakic, Thank you so much for the feedback and for the Grishin reference, I will definitely look into his work. I completely understand not having the time to read through a full preprint

And you are right to be skeptical, I was too. A result like that could easily be a statistical fluke. But the thing is, I found a similar coincidence when I was looking at the sweet-tasting proteins Thaumatin and Monellin, which are also classic examples of convergent evolution. In that system, I found two separate "sockets" with no known function. One of them, a two-residue segment (Thaumatin 157-158 Chain A vs. Monellin 30-31 Chain A), aligned with an RMSD of just 0.049 A. finding two of these statistically improbable geometric echoes in completely unrelated systems is what got me thinking about this whole "Geometric Deep Homology" idea. It's a very curious puzzle. Thanks again for the discussion!

https://zenodo.org/records/15926235

[Mensur Dlakic]

in the reference GenoMax provided, the abstract says this:

Structure determination of the catalytic domains of two members of the kinesin superfamily reveals that this class of molecular motor exhibits the same architecture as myosin and suggests that these microtubule- and actin-based motors arose from a common ancestor.

Like I said before, your original point about the classic case of convergent evolution between myosin and kinesin seems to be flawed. I assume the same might be true for your other example of the sweet-tasting proteins.

By the way, it doesn't strike me as particularly remarkable that the two consecutive residues would align to some other two consecutive residues with RMSD=0.049 A, especially when this similarity is found in functionally related proteins. Still, I would assume this happens in a relatively small number of cases.

[Alberto]

Dear Mensur Dlakic, Thank you for your continued insightful feedback. I particularly appreciate your critical perspective on the uniqueness of these findings. I agree that functional relatedness could indeed contribute to some level of structural convergence, as you suggest. It's a valid point that merits further exploration. My personal interest in this area stems from a broader hypothesis that these isolated pockets of sub-Angstrom geometric identity might represent a deeper principle i'm trying to understand, I'm currently exploring other systems to see if this pattern holds. Thank you again for your time and willingness to discuss these preliminary observations.

[GenoMax]

Monellin and Thaumatin are of significantly different size and show no sequence homology. They have completely different tertiary structures and belong to distinct protein folds. So they show functional convergence. (Ref: https://pubmed.ncbi.nlm.nih.gov/3614382/ )