11 replies [Last post]
S0ckrates's picture
User offline. Last seen 22 hours 58 min ago. Offline
Joined: 05/19/2017
Groups: None

Last night on stream I had noticed that whenever I accidentally created a Michael Acceptor in my molecule my score shot down at least 6000 points or so. Is there an undocumented toxicophore detector in the SMD puzzles? If so that's super neat; I was just thinking about how it might be easy to overlook a known toxicophoric structure in hand-modification of lead molecules in my computational drug design class.

Joined: 12/27/2012
Groups: Beta Folders
check the conditions...

I ended up with a "likely unsynthesizable" penalty, -6000 points, when I tried to build Darunavir, (DRV, PDB O17).

Unfortunately, I wasn't watching the score, so I'm not sure what part caused the problem.

Darunavir has two stacked rings at one end, kind of halfway to a Bucky ball, and you can actually build it in Foldit.

S0ckrates's picture
User offline. Last seen 22 hours 58 min ago. Offline
Joined: 05/19/2017
Groups: None
See that's the thing...

I didn't see an unsynthesizable penalty in my conditions tab. See below. I have the structure of a Michael Acceptor from a slide in one of my classes, and you can see I've replicated one in my molecule inadvertently with the amine group. For comparison, I've also uploaded a screenshot of my best where the double bond is fully substituted, and there isn't a second electron withdrawing group (the ketone); the mechanism for the Michael reaction cannot proceed, and the toxicophore is rendered harmless. As you can see, no changes in the filter for synthesizability. And yet, 3000 points. I'm willing to bet that's not from conformational analysis, and in the score breakdown, it seems to come from the Ideality subsection.

Joined: 09/22/2017
Groups: Beta Folders
This is a fantastic web

This is a fantastic web resource on reducing problems with pre-clinical drug candidates: https://www.cambridgemedchemconsulting.com/resources/ Might I recommend cyclizing that cyclopentane ketone to the nitrogen making a heterocyclic ring system?

S0ckrates's picture
User offline. Last seen 22 hours 58 min ago. Offline
Joined: 05/19/2017
Groups: None
Losing out on H-bonds

Unfortunately I was taking advantage of the twin H-bond donors of the amine and I just tried that and losing out on it results in too much of a score loss. Great resource though, I was actually building semi-blindly from scratch this time around and if I was gonna take the cookie cutter rational drug design approach I would certainly use resources like those; I just wanted to see how far I could get with raw intuition and in-game cues, as is my style with what a lot of Foldit puzzles actually.

jeff101's picture
User offline. Last seen 16 hours 16 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
What's a toxicophore?

https://en.wikipedia.org/wiki/Toxicophore says:

A toxicophore is a chemical structure or a portion of a structure (e.g., a functional group) that is related to the toxic properties of a chemical. Toxicophores can act directly (e.g., dioxins) or can require metabolic activation (e.g., tobacco-specific nitrosamines).

rmoretti's picture
User offline. Last seen 1 hour 32 min ago. Offline
Joined: 01/15/2010
Groups: None
No toxicophore detection currently.

We don't have toxicophore or "bad substructure" tests present in the small molecule design puzzles yet. It's something we're looking into, but nothing that's active yet. (We're hoping to get some more information about the sorts of edge-cases Foldit players like to come up with, first - e.g. what's the "megahelix" for small molecule design.)

If the score penalty is coming from the "Ideality" subsection of the score, then the issue is not the toxicophore, but internal strain in the geometry of the ligand. While a global wiggle should reduce this somewhat, a ligand-specific wiggle can do a much better job. If you select just the ligand and do a wiggle (either in selection mode or through the right-click pie menu in original mode), it should do a slightly different wiggle approach which should do a better job in removing those ideality issues for designed ligands.

The MMFF wiggle (either in selection mode or from the right-click pie menu) should also greatly help with ideality issues, though as it's working with a different energy function than the normal wiggle, there's a chance an MMFF wiggle will decrease your score slightly.

S0ckrates's picture
User offline. Last seen 22 hours 58 min ago. Offline
Joined: 05/19/2017
Groups: None
Tried this last night

It wasn't immediately obvious, but this actually really helped out with score gains. Thanks for the clarification.

jeff101's picture
User offline. Last seen 16 hours 16 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
MMFF questions:

I found the MMFF wiggle button in the selection interface.
It appeared as a button labeled "M" when I selected just
the ligand. When I selected the ligand plus parts of the
protein, this button disappeared. Is it possible to do
MMFF wiggle on more than just the ligand?

In the original interface, why not put the MMFF wiggle
button in the Actions menu along with all the other
wiggle buttons?

Are there hot-keys or LUA functions for MMFF wiggle?

rmoretti's picture
User offline. Last seen 1 hour 32 min ago. Offline
Joined: 01/15/2010
Groups: None
Mostly ...

The MMFF wiggle does take into account the surrounding protein structure, but as a fixed entity, rather than something it can optimize. As mentioned, the MMFF scoring is really intended for small molecules, so it doesn't necessarily do all that well when optimizing proteins or protein-small molecule interactions. That's why we don't have it on generally.

In the most recent client, the MMFF wiggle should be available in the right-click pie menu. Again, because it's ligand specific, it's closer to a local wiggle or the other local actions that are in the pie menu, rather than a global action like is in the main actions menu.

At the moment none of the small molecule design functionality is exposed at the LUA level, and the hot keys are minimal. In selection mode pressing "M" should do an MMFF wiggle if the ligand (and only the ligand) is selected.

jeff101's picture
User offline. Last seen 16 hours 16 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Lipinski's Rule of 5:

I like S0ckrates' tips
(like Lipinski's Rule of 5 in #18)
for designing small molecules at:
https://fold.it/portal/node/2005231#comment-36784

Perhaps something like Lipinski's Rule of 5
could be built into Foldit.

S0ckrates's picture
User offline. Last seen 22 hours 58 min ago. Offline
Joined: 05/19/2017
Groups: None
It already is! Kind of.

The game is already checking for logP values, HBA/HBD counts and molecular weight in the conditions for 1519! Just, not as strictly per se. There's a (pardon the pun) bit of wiggle room before the game starts docking off points.

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons