JM: Sure. Just have a couple of plots to show. So this is from the benchmarking of e-molecules.
For the wiberg bondorder, so this is for like a system with two rings and the bond in between that. So this has a lot of data on it. But as you can see, the distribution isn't that nice.
DM: Yeah, it's good to see that molecule count finally get up there where it should be, though.
JM: Yeah.
DM: Okay. I was going to say, what if you put it on a log scale? But really, I guess it gets out and this is the upper limit on the right side, unless you've done something very with the bars, like you're seeing nothing above like 1.03 or something, right.
JM: Yeah. No.
DM: Yeah. No need to change the access scale or something. Remind me, like what we wanted to assert. Like, I remember discussing that, and I don't remember what we were trying to look at. Oh, yeah. So the improper torsion, basically. So I think the point, Jessica, is that like in general, when you're pulling the WBOs, it would be good to pull the torsion at the same time. But you probably don't need to bother doing that here. Going back to it here. So basically just when you when you're processing the data in the future, try to grab the torsion angle at the same time. But no need to backtrack.
JM: The next plot is looking at the average improper angle versus the wiberg bond order. Um. The correlation is not so nice.
I made an interactive plot with that too, but it's a little overwhelming. So I guess my next ideas were to. Well, I don't know if this kind of seems like it's not so promising for WBO, like, interpolation. But like, my next idea was maybe. Try to look at, try doing the same plot with all of the existing and properties in the force field and see maybe some of these parameters would be good for interpolation.
DM: Yeah. The existing impropers I think most of them are very, very broad.
JM: Yeah. They're fairly broad, I guess. I can show them.
DM: Okay.
JM: So these are the current ones.
DM: Yeah. Okay. So there is any thing connected to a carbon, which is double bonded to a nitrogen, and then connected to another nitrogen. That's the bottom. So that one's not super general. What's the one next to the bottom? That is very weird looking.
JM: Yeah.
DM: Okay. That's a generic one, but involving a nitrogen in a bunch of different situations, I guess. I mean, I don't have a problem with running that experiment. I think that some of these it's unlikely they'll show anything but some, especially that bottom one might be where it might might there might be a chance. The other question I have here is on this relating to this graph. So can you flip back to Slide 14 for a minute? Okay.
So that goes from 1.0 to 1 point. A little over 1.2, but most of the action happens between 1.0 and 1.15. So if you go back now. So, this covers enough range. It's not a problem with not finding enough range. Do you want to go to your interactive plot for a minute? Okay, cool. So let's take the molecule that's sort of on the top right?
Yeah. Interesting. That's really seriously congested. Can you go to like one of the molecules on the top left now? Anything in there? Well, we. Okay. Yeah. Let's look at some more of those molecules. Keep going. Yes. More. Yeah. Keep going even more and then just move down the plot a ways. All right, cool.
DM: Yeah. I guess when I see the molecules, I suppose I'm not that surprised. We're not seeing anything here because it kind of seems like. Know, I'm interested in a lot of things, but it kind of seems like we're just running into the same thing that's made the distortions so difficult to understand in the. These are big, complicated molecules with all kinds of things bumping into each other, and that ends up dominating the energetics rather than the bond being an important factor, for one. Yeah. So. And what's the data we're looking at in this part? This is like all of QC archive data or something.
JM: So you have all of the matches in the optimization data sets.
DM: Okay. Painful in the sense of why it's not obvious what we can do to push this forward without. I was hoping we would see some low hanging fruit or something that would obviously. It's really we can make this work, but it looks like there's not an obvious answer.
JM: Maybe another idea is trying to filter the molecules like by some sort of size, because maybe the larger they are, the more like interactions they're having. And it's kind of convoluted in the relationship between the WBO and the improper angle. I'm not really sure all of these things. I haven't looked at the sizes. Size distribution.
DM: Yeah, I mean, that's interesting. The part of the issue probably is not just size like over here, here you have a methyl on the N20, which is going to clash with the adjoining aromatic ring and be a really important player in what the geometry of that is. So you could have that methane there even if it were a much smaller molecule. But that said, that's probably important chemistry. And probably ought to tell us something about the energetics of that improper. So maybe that's not a bad thing. Maybe it's maybe here. Part of the issue is, well, there's this whole other bicycle ring on the other end that's also coming into play, so I don't know.
JM: Is there a way to filter like determine steric interactions on a specific atom and then. Just I'm not sure if it exists.
DM: That's a good question, I was wondering the same thing. So. It's hard to compute absolute steric interactions because molecules have internal energy interactions. What we were doing for the torsions, we were looking at how much the Leonard Jones energy varied as you drove the torsion and if it became very, very unfavourable, then we saw that as a strong steric clash. These are opidization sets, so we don't have like a torsion profile. Is it clear why this is difficult, Jessica, why it is difficult to check the sterics?
JM: It's just because there isn't really a way to measure that.
DM: Or just end up with like a total energy for the energy component. And it includes like attractive and repulsive interactions between lots of different atoms. I mean, you could zero out some of them, I guess, and then you forget you can decompose it into specific interactions. But still, since you don't know which items are you're worried about clashing, you don't think you know which items to zero out if you're trying to estimate it. So it's quite possible. Like, for example, you can have a molecule that has a favorable energy, even though it has a relatively strong steric clash somewhere because the favorable parts in one, the favorable interactions in one part offset the undesirable aspect in another part. But the steric compulsions are the stiffest interactions. So it may be that you would tend to see molecules with more the steric clashes have less favorable energies anyway, so it may be worth a quick check.
JM: Okay.
DM: I think you know, I don't think it wouldn’t hurt to just pull the Lennard Jones' energies. Well, the easiest way to do it would be if you know how to overlay the notation into these pop up plots, just embed the LJ component of the energy. Respond to these images and then you can scan through the plot and see if it looks like things with stronger steric issues have less favorable LJ energy. Or you could color code your plot by them or something.
JM: Can you calculate that with the open source toolkit?
DM: So with openmm you parametrize a molecule, then you do an energy evaluation and you pull the components of energy. I expect that there's something Pavan probably has an example of that. You have an example handy for one. Yes, I know I've done it before, but it's been a year or more so. You would look for cases where the difference in energy between conformers is really big. Yeah the ones I think that are going to be the most are the ones where you can't really get away from this interaction. So like in the case you're looking at a minute ago, you have a ring with a methane coming off of it and that methane can't avoid bumping into this other ring. And so all your conformers are going to have a steric clash. But. It is true that molecules with big differences between conformers are going to tend to be bigger and fluffier with more steric issues, probably. So it might not be a bad thing to look at. It just may not catch all of the cases. Okay. So, Jessica, why don't you just take a look at that energy and then as soon as you have that, we may be close to a point where we say let's stop going in this direction because it doesn't look like it's getting us anywhere. So we may want to talk again before a week from now. But we should look at that first.
JM: Okay.
DM: So make sense to you.
JM: Yeah. Sounds good.
DM: Did you have other stuff ohh its your last slide. Okay, thanks for taking out all this data. I'm glad to finally see those molecule counts go way up. Maybe you want to update Chaya on with these two slides, just by putting an update in the slack or something so she can think about it a little more. I'm disappointed that we don't see a big range in every bond or when we look at all of the molecules now. But it is what it is.
JM: Yeah.
