[FRIAM] Schwill Rock?

[Marcus Daniels]

A search algorithm that, say, proposes a prefix or a suffix to a SMILES string would need to have a way to autocomplete candidates before it could use these descriptors to guide an optimization because the parsing step is non-trivial, never mind the sanitization step (mentioned on that web page).

I will deflect blame on Jon for changing the topic from music to chemicals, but presumably with enough debate their aesthetic preferences in music could be formalized in theory or some rule-based way, as is manifest here.

From: Friam <friam-bounces at redfish.com> On Behalf Of Roger Critchlow
Sent: Tuesday, October 12, 2021 2:10 PM
To: The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
Subject: Re: [FRIAM] Schwill Rock?

Hmm, when I was in the drug discovery canal, the "descriptors" that you could calculate from a SMILES string were legion.

Here's the list for RDKIT, https://www.rdkit.org/docs/GettingStartedInPython.html#list-of-available-descriptors.  There are one bunch that depend entirely on the formula and molecular structure.  Then there's a whole other bunch you can compute if you generate 3d structures for the molecules, possibly multiplied by the number of low energy structures the molecule can adopt.

What kind of plausibility were you looking for?  Does the SMILES string specify a real molecule?  That's hard.  There are syntax errors in SMILES, failures to close rings, valency errors, charge errors.  But there are lots of syntactically valid SMILES that won't match any known molecule, either because they're impossible or as yet to be determined.  The pharmas all have their own lists of molecules of interest, but those are proprietary.  Looks like there are various online databases, none that I'm familiar with.  If the SMILES parses, you can try generating a 2d depiction and a 3d structure.  Those will throw exceptions if things get too weird.

-- rec --

On Tue, Oct 12, 2021 at 3:22 PM Marcus Daniels <marcus at snoutfarm.com<mailto:marcus at snoutfarm.com>> wrote:
I was playing with RDKIT the other day, and it wasn’t obvious how to get a scalar quantity of plausibility of a molecule.   It seems a SMILES string is right or wrong, and then maybe there are some warnings that can be trapped.   However, the benefits for search or fair sampling are different than the needs of correctness checks, which is deeper property.   That isn’t quite a fit to the music example where aesthetic considerations are subjective.

From: Friam <friam-bounces at redfish.com<mailto:friam-bounces at redfish.com>> On Behalf Of Jon Zingale
Sent: Tuesday, October 12, 2021 12:11 PM
To: friam at redfish.com<mailto:friam at redfish.com>
Subject: Re: [FRIAM] Schwill Rock?

"I mean from the perspective of aesthetics. Understanding why Pandora is messing it up means sampling the deep wells."

Yes, but not more than one has to. This is why I am advocating for methods like a weighted ensemble. The working analogy for me comes from drug discovery. It doesn't make a lot of sense to probe the same old sites and the same old conformations.

.-- .- -. - / .- -.-. - .. --- -. ..--.. / -.-. --- -. .--- ..- --. .- - .
FRIAM Applied Complexity Group listserv
Zoom Fridays 9:30a-12p Mtn UTC-6  bit.ly/virtualfriam<http://bit.ly/virtualfriam>
un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives:
 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
 1/2003 thru 6/2021  http://friam.383.s1.nabble.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20211012/2bd14aa5/attachment.html>