jchodera
commented
1 year ago These are good ideas!
Here are some other ideas, some of which have been mentioned before but are not listed above:
-
Enamine building blocks. These are chemical "building blocks" used in reactions at Enamine or elsewhere for synthetic organic chemistry for drug discovery. We can download the SMILES libraries, and if there is any concern in securing permission to use these compound libraries, we can ask for that (we work with Enamine closely). While PubChem provides coverage of molecules that have already been made, good coverage of building blocks provides coverage of molecules likely to be made in the future. If size is an issue, it could be filtered by size and/or clustered by similarity.
-
Enamine REALSpace representative couplings. These would be a small subset of representative couplings between small building blocks to provide coverage of the chemistries joining synthetic building blocks together. These could be generated via common robust SMIRKS reaction transformations that couple together compatible SMILES for building blocks.
-
Common post-translational modifications of proteins. A large number of post-translational modifications are commonly found for proteins, so confining ourselves to natural amino acids is severely limiting for studying real biomolecular systems.
-
The PDB Chemical Component Dictionary. This is a catch-all of everything anyone has cared enough about to obtain a structure. There are statistics available for how many times each entity appears that can provide an easy way to capture the smallest number of most important residues.
-
Lipids. Lipids are important for membranes and membrane proteins, but there poor coverage of these in quantum chemical datasets. There are databases of lipids and lipid structures that could be tapped. Lipid nanoparticles are also incredibly important right now.
-
Carbohydrates. Complex carbohydrates are attached to many eukaryotic proteins, and are composed of a relatively small number of building blocks and common linkages. A representative set can readily be generated if someone hasn't done this already.
-
Nucleic acids. Nucleic acid coverage is important for simulating RNA and DNA, which are both important for understanding biomolecular mechanisms and also targets for new technologies like RNA vaccines, CRISPR, and targeting RNA with small molecules.
How should we prioritize the various ideas? Is there a simple voting scheme that could be used to capture interest? Or is it better to focus on an MVP that delivers clear value to many users.
-
One compelling MVP is "being able to simulate most of the PDB". Since structural information is required for molecular simulations and the PDB is the primary source of structural data, tackling the most popular chemical entities from the PDB Chemical Component Dictionary (and perhaps common couplings they are involved in) would be extremely high value.
-
Much of the drug discovery world now uses the 38B+ compound Enamine REALSpace one-step synthetic library for early stage discovery, so delivering greater coverage and accuracy for these compounds would be very high value.
-
Nucleic acid, lipid, and carbohydrate force fields are critical for representing many biomolecular systems that are difficult to simulate right now. Providing greater coverage of these with targeted datasets would be highly valuable.
peastman
pavankum
jthorton
giadefa
JoshRackers
I've been talking with lots of people about what would be most useful to add in version 2 of SPICE. I've gotten lots of great suggestions. Here are some ideas in no particular order.
Solvated PubChem Molecules. This would provide data on ligand-water interactions. They could be generated the same way we did the solvated amino acids: simulate the molecule in a box of water, periodically saving conformations along with ~20 of the nearest water molecules. These will be largish, and therefore expensive to compute. We'll probably need to limit the number of molecules we include and the number of conformations for each one. I think this is a case where a little data can go a long way, since each conformation has many water molecules.
Water Clusters. This would provide some conformations that are more representative of bulk water. How large should the clusters be? What's the best way to generate them? I've been advised that classical water models often don't produce very realistic water structure.
PubChem Molecules with Boron (and possibly Silicon). I excluded these elements from the original PubChem set. They only appear in a tiny fraction of molecules, so there wouldn't have been enough data on them to be useful. But if we specifically select molecules that include them, I think we could assemble a useful amount of data. There's around 1500 molecules with B and around 1900 with Si.
Amino Acid / Small Molecule Pairs. This would provide more data on protein-ligand interactions. One way of generating them would be to use PDB structures. We could identify which amino acids contact the ligand and generate conformations for them, each conformation including the ligand plus a single amino acid.
Transition States. If you're interested in rates, it's important to model transition states accurately. Because they're higher energy, they won't be very well sampled in a thermal distribution. Can we add conformations specifically to sample transition states? One suggestion was to scan dihedral rotations while minimizing all other degrees of freedom. What molecules would it be useful to do this with? Of course, not all transition states involve dihedral rotations. What else could we do?
Metals. The goal would be to allow simulating metalloproteins. These could also be generated from PDB structures, taking the amino acids that interact with the metal. Would it be sufficient to consider only one amino acid at a time, or do you really need to include the whole environment? If the latter, these could be quite expensive to calculate.
Active Learning. This is the idea discussed in #35, to add more conformations for existing molecules to improve the accuracy of trained models.
More PubChem Molecules. More chemical diversity is always good. The number included in version 1 was mostly arbitrary, just based on how long it took to do the calculations. It would be easy to add more.