What is the goal of small molecule/ligand

Case number:671071-2004240
Topic:Game: Other
Opened by:alcor29
Status:Open
Type:Question
Opened on:Monday, September 25, 2017 - 20:57
Last modified:Monday, September 25, 2017 - 23:40

From what I've seen on 1432 there is a small finite number of atoms we can use to design the "ligand", a small number of positions that can be occupied by these atoms and a small number of angles which we can select. This seems to be a task more perfectly suited to a digital treatment. Have the program multiple all the probabilities together and then run a basic stability test on them. That would seem to leave then the task of figuring out how the protein can fold around it. And that is what have been doing all along isn't it? Dealing with the more infinite questions of how the structures are oriented to each other in 3D space. Please help me understand then what value there is then in having us swap atoms, positions and angles.

(Mon, 09/25/2017 - 20:57  |  4 comments)


rmoretti's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 01/15/2010
Groups: Foldit Staff

The issue with doing a fixed enumeration is the combinatorial complexity. For a drug-like small molecule, it's been estimated that the number of possibilities is 10^60. To put that in perspective, if a 3.5 GHz processor was able to make and test one molecule per cycle (absurdly fast: it normally takes millions to billions of processor cycles in order to test a compound), it would take about a trillion, trillion, billion processors all working for the entire history of the universe (13.7 billion years) in order to enumerate all of them. - So while at any one place there might be only a small number of possibilities, those small numbers add up (or rather, they multiply up, which is what causes the issue).

Just like protein folding, we can't hope to run through all the possibilities in any reasonable time. That's where Foldit players come in. Most of the compounds in that 10^60 aren't going to be useful, and obviously so. If Foldit players can recognize how small molecules interact with the protein, they can use their intuition to figure out which atoms in which locations are useful for making good interactions with the protein, and only try to build those.

So that's the ultimate goal of the small molecule design puzzles. We hope that - just like how the community has learned how to best fold proteins - they can learn how small molecules interact with proteins well enough that they can propose interesting new small molecules which can be tested.

alcor29's picture
User offline. Last seen 1 day 1 hour ago. Offline
Joined: 11/16/2012

Thanks for the quick response Moretti. I guess I was only seeing the multiplication of the three variables I was seeing as tools on the application. And the multiple didn't seem to be more than 4 orders of magnitude; nowhere near as high as 10^60. Are you including the ligand-protein interactions? If you are then I still have the question about the ligand design I stated above. If you are not, it would be interesting to know what other variables are involved in ligand construction without considering protein interactions.

rmoretti's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 01/15/2010
Groups: Foldit Staff

The 10^60 estimate is just for the identity of the small molecule, not the interactions with the protein.

One way to think about it is to think about each atom in a molecule. If you choose between just three elements (carbon, nitrogen, oxygen), and there's 20 C/N/O atoms in the molecule, you have 3 billion possibilities already (3^20 being all the different ways you can pick from C/N/O at each of the 20 positions). But that's just varying the element. You can also vary the number of atoms and how they're bonded. (For example, there's at least five places you can add an atom off of a 6-atom ring attached to a molecule.) How this affects the calculation is more complicated to calculate, but it likewise blows up multiplicatively. Researchers in the late 90s made some reasonable assumptions about how to calculate things (including how to ignore nonsense molecules and duplicates) and got the 10^60 estimate out.

alcor29's picture
User offline. Last seen 1 day 1 hour ago. Offline
Joined: 11/16/2012

PS. I did not vote down your answer. I meant to click reply but my hand slipped.

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