openfreeenergy / openfe_analysis Goto Github PK

Type
FE calculation check
Question
What estimate would I have produced at different points in the simulation?
How
Time series of estimates
Tool
PyMBAR

Type
System sampling check
Question
Still in complex?
Tool
MDA

Type
System sampling check
Tool
MDA

Thinking of this as a user story - if I have a host guest system I won't have a protein but I'll probably want to still align based on the non alchemical host.

My understanding of https://github.com/OpenFreeEnergy/openfe_analysis/blob/main/src/openfe_analysis/rmsd.py#L40 is that this wouldn't work right now?

In the Tnks2 system from the protein ligand benchmark set, the ligand RMSD, ligand COM drift and protein RMSD are likely not calculated properly. The protein consists of two chains which may have led to PBC not being removed properly and/or wrong alignment and therefore these very large RMSD values.

Protein and ligands:
https://github.com/OpenFreeEnergy/protein-ligand-benchmark/tree/changes_plb/data/tnks2

Ligand COM drift, Ligand RMSD and Protein 2D RMSD:

Type
System sampling check
Question
Are there poorly sampled changes in binding site side chain rotations that aren't being sampled across some states?
Tool
MDA
Details
Rotamer detection

Type
Enhanced sampling check
Question
Enhanced sampling boost works?
How
All replicas at least one round trip
Note
D.Sidler did define this nicely (also round trip time etc.)
Publications
Efficient Round-Trip Time Optimization for Replica-Exchange Enveloping Distribution Sampling (RE-EDS) - PubMed (nih.gov)

Type
System sampling check
Question
Is phase space well sampled?
How
Minima sampled ~60%
Tool
MDA + rdkit
Note
Ideally what you should be looking at here is a combination of the solvent, vacuum, and complex differences and the variation in dihedral sampling across lambda states. I.e. is there a major change happening in my transformation that isn't being sampled across all states -- @IAlibay

Type
System sampling Check
Question
Still in complex?
Hard to get convergence?
Tool
MDA

Type
System sampling check
Question
How does the thing look?
How?
gif of simulation/some other easy video format
Tool
image magick?
Note
Med chemists like this
Details
Could be embeddable HTML of single frame

Type
system sampling check
Question
Still in complex?
How
RMSD < ~5 A
Tool
MDA
Note
Size dependent

Lack of symmetry is ligand RMSD is rather problematic, using something like https://spyrmsd.readthedocs.io/en/develop/ would be better but doesn't have an easy solution for a) PDB-only inputs, b) dealing with alchemical molecules.

Type
System sampling check
Question
Look for binding waters and their exchange

Type
System sampling check
Question
Is there an ion walking into my ligand?
Tool
MDA?

Type
FE Calculation check
Question
Is the simulation approach converged? (from sim perspective)
How
Stationary result for a given time
Tool
PyMBAR
Publications
A Simple Method for Automated Equilibration Detection in Molecular Simulations | Journal of Chemical Theory and Computation (acs.org)

Type
Question
Does binding site have an impact on the dG
Publications
Zhang et al

Type
FE calculation check
Question
Can the FE estimate be accurate?
How
Dist overlap >1-5%?
Tool
PyMBAR
Note
Koenig & Boresch published on this

Type
Convergence check
Question
Have simulations converged?
How big is the cycle closure error?
Which leg in the network has not converged yet?
How
Sum ddGs
Tool
networkx

Type
System Sampling Check
Question
Protein still folded?
Tool
MDA
How
RMSD < 6 - 10 A
Note
Size dependent

Type
Setup check
Question
What FF was applied to my model?
Tool
OpenFF + RDKit
Details
For each atom in SMC -> epsilon + sigma + charge assigned, all bonded terms

Type
System Sampling Check
Tool
Prolif
MD-SIMBA