2021-04-09 WBO/Impropers meeting notes

Participants

@Christopher Bayly

@Hyesu Jang

@Simon Boothroyd

@Jessica Maat (Deactivated)

@Trevor Gokey

@David Mobley

@Pavan Behara

Discussion

PB: I used Simon’s constructure package to build a set of biaryl molecules for the bridging bond torsion without any steric effects from ortho substituents and spanning a wide range of wbo values. Most of the molecules fall in the range of 1.0 to 1.01 and some molecules till 1.15 and very few beyond that. Here’s a summary of work.

CB: Okay, it would be good to do constrained optimizations with the torsion angle set at 0 and 90 along with an unconstrained geometry optimization. The unconstrained geometry optimization would give us the dihedral values around 40 + 5 for wbo of around 1 and it would be planar at higher double-bondish wbo values. Doing this would give us a rough idea of how good is wbo interpolation.

SB: Yeah, sounds good. Just a reminder that QCA queue would be busy for the next two months or so with Sage fitting datasets, so for now let’s wait for a while before jumping on this.

PB: Okay, will get to this later.

HJ: I want to show my torsion dataset creation procedure for Sage fitting, main purpose if to reduce contamination from sterics and other sources. I am joining two substituents and rotating that bond and I made a list of around 360+ substituents from Roche and other datasets. With around 60K molecules I am picking up atleast 20 or so using various methods like maccs similarities, etc., bringing the total torsiondrives to nearly 3000. Any feedback is appreciated on the method of selection, molecule building, etc.

CB: This is excellent work HJ. One suggestion I have is to pick molecules based on best-first heuristic, in the sense pick more pharmaceutically relevant and most frequently encountered rather than some artificial molecules like a pyridyl notrogen attached to a chlorine etc. Few sources of such data are Enamine, ChemBL, e-molecules, etc. After generating the dataset may be a check against these databases would be great.

HJ&SB: Sounds good, will try to work on that. Although some of the molecules are artificial we think the electron density and the chemical diversity should be good enough to inform the FF.

DLM & CB: May be send this set of substituents and ask pharma partners to rank them.

SB: We are on a tight timeline and it would be a huge ask.

DLM: Then may be just a quick glance and root out not so relevant substituents would be great. HJ can post in slack and I can review and tag pharma partners to give a quick feedback.

HJ: Sounds good, will do that.