The InChI normalization step may also change the chemistry. For example, it may put things into a preferred tautomer form or charge state. This is essential for its goal of linking disparate records, but it is not chemistry preserving, and can be worse than passing an SD file to your nanomachine factory.
It was designed as "deterministic ID for sharing", yes, but not for sharing structure information but rather information related to the structure.
The most "concise" solution is to record the positions of the nuclei and the total charge. Then any quantum mechanics program will be able to reconstruct (to the desired level of approximation) the electron density of the molecule. This is the one you reject.
For good reasons. As you say, you want a nomenclature that can help with identity and similarity, and electron density is a very slow way to do that.
But such nomenclature is driven by human concepts. Fundamentally, there are different reaction rates depending on the relative orientation of two molecules. We see this in molecular beam experiments. For example, from 1979, http://www.sciencedirect.com/science/article/pii/00092614798... "Molecular beam reaction of K atoms with sideways oriented CF3I".
InChI doesn't have a special name for "sideways oriented CF3I", even though a complete, concise system would have to be able to provide a name for it.
Instead, we've come up with concise approximations - a SMILES string, a bitstring fingerprint, etc. - which are tractable for the purposes you want, so long as we stay within relatively normal chemistry. But it's still humans who categorize things into, say, "aromatic" or "chiral", so our approximate nomenclature has to partially encode how humans think of chemistry.
The Game of Life is much simpler than the Reality of Life.
Though the Game of Life is surely simpler than the Reality of Life... if one were to flat out name the core concept as "godel numbers for chemical formations", it's certainly possible, don't you think? InChI isn't it, I agree; just wanted to toss the seed of the idea out there, with an example that's a lot "closer" to the target semiotics than a bunch of unnormalized floating point coordinates drifting around in xml.
"good enough is the enemy of good" can apply in spades here, so I appreciate your thorough analysis of specific weak points in InChI.
InChI is only closer for the classical purpose of chemical documentation - record association and text-based search. That is not a bad thing, but given the entirety of your goals, it is equally correct to use "close" for SMILES, SLN, SD, Mol2, etc.
This is the slowest spaceship known, right? (Which isn't a machine made out of other parts?)
It also helps that Games of Life can be saved and reloaded. If your lab bench explodes, too bad, start over from scratch (assuming you're uninjured).
(cough)
Are we sure of that yet?
> Real Life physics is a little less apt to collapse into a seething mass of undifferentiated matter of you poke it in the wrong place.
Are we sure of that yet?
RL follows quantum mechanics. As far as we know, its state cannot be recorded perfectly.
GoL is a Turing Machine. Its state can be recorded perfectly.
RL follows relativity. There is no equivalent in GoL.
Hashlife gives a method to compute GoL boards faster than processing each cell, and gives an exact answer. There is no equivalent for RL. At best there are probabilistic predictions. (There might be a 1E-50000 chance that the Sun won't rise tomorrow because all of the atoms in the Sun fuse at the same time and it blows up.)
Of course there is an equivalent of relativity! There's a fixed, ultimate speed-limit (i.e. c) of once-cell/time-step. It is not possible to transmit information faster than this in GoL.
https://en.wikipedia.org/wiki/Superfluid_vacuum_theory#Relat...
"Relativistic quantum field theory"
This is already experimentally observed in our own universe. Quantum noise exists in vacuums.
If you look at C in GoL as the fastest an actual black square can propagate, then of course you'll validate your naive assumptions that our universe could never be digitally replicated by GoL. But if you take the idea of implementing a primordial sea of quantum noise, and looking at the GoL board not as physical particles directly, but as a deeper construct. Then you can actually replicate relativity.
To estimate probability of an event you must have a sample were both have occurred, to provide a probability for something like the Sun not rising tomorrow is pure non-sense.
The same can't be said for the atoms in the sun.
You also have to make an arbitrary decision about when the cannonball touches the ground. How much interaction is needed between the atoms of the cannonball and ground before you say there's contact?
These issues don't exist in the GoL universe.
Outside the universe ain't. But even though, I can confidently say that we are definitely inside.
Not a very tangible idea, but not useless of purpose of reasoning about the affairs of things.
tl;dr: easier problem is easier!
Chemistry is dealing with a lot, lot more of concepts than the study of the Game of Life. Brevity only works when you have few enough information that you can map it to only concise words.
(Entropy strikes again, even if it's information entropy for a change.)
Information entropy and physics entropy are actually the same thing. You can measure heat capacity in bits.
[1] https://en.wikipedia.org/wiki/Simplified_molecular-input_lin...
SMILES is better at #1 than InChI, but not as good as a 3D structure representation. But the 3D structure isn't as compact, and offers few advantages for structure identity for small molecules where the 3D configuration is more easily predicted from geometry.
InChI is better for #2 than SMILES because different tools generate a different canonical structure, and because general purpose text search engines like Google don't do so well with SMILES as InChI keys.
Fingerprints, either binary or count, are more often used for the similarity aspect of #3, than n-gram solutions using the SMILES directly (like LINGO), or any method based on InChI. Also common are 3D shape and electrostatics comparison methods, with attempts to sample configuration space.
And for certain chemical structures other approximations are even more important. DNA similarity is based on very different principles than used for small molecules.
Very close, yes. But if we compare to, say, reproducible builds (because it's also dear to my heart), the consequences of that difference are legion. For example: the `tar` command can definitely extract you a filesystem, yes, absolutely. However: whether or not two `tar` implementations (or even the same implementation, run twice in a row) produce the same byte stream is make-or-break for whether someone else can follow your procedure and get the same result.
If the process is deterministic, the identity check is cheap: hash it. If the process requires unpacking again to do semantic checks on the contents, the amount of time is... well, if we're hand-waving on the format, essentially unbounded.
So... yeah, "canonical SMILES" is a valid answer; that "canonical" part is just really important.
For a variation on the theme, if we all agreed on a canonical SMILES method, then the SHA256 of that could be used as a key to connect records, without revealing what the underlying structure is.
This is done now, with InChI Key, when two companies want to compare overlap without sharing information about internal compounds. (There was a paper about this last year or so, but I can't find it.)
I also pointed out that InChI Keys are better for text-based search engines. As an extreme example, I could produce a line notation which uses only the characters "(", ")", ".", and "-". It would be fine for #1, but poor for #2 for any database search engine which expects some concept of "words" made of letters.
This is why I distinguish #1 from #2, even though they are indeed close.
Now if only we had descriptions of chemistry that were this terse. Imagine if this kind of problem solving, collaboration, simulation, and instant verification were the norm for synthetic chem. One of the comments -- "[Let's] use gencols to rub the ship against gliders and *WSSs to see whether there is a useful collision to maybe build a puffer" -- just blew me away. If this were chemistry, that commentator would have been suggesting automatic nanomachine factory discovery.
(InChI appears to be close. But vast amounts of data are locked up in obtuse formats are either Assigned-Names-And-Numbers style formats which are useless to indexing and similarity searches, or formats that embed non-relative coordinates in 3d space, etc, in such a way that computing a deterministic ID for sharing is practically a nonstarter.)