Add molecules that provide coverage of post-translational modifications of proteins

[jchodera]

We might consider adding a molecule set that provides coverage of post-translational modifications of proteins (e.g. phosphorylation of Ser/Tyr/Thr residues, methylation of lysines, etc.).

One way to do this---which might also provide coverage of nonnatural amino acids---is to extract a subset of the PDB Chemical Components Dictionary that matches the SMARTS pattern for an amino acid backbone (e.g. [NX3,NX4+][CX4H]([*])[CX3](=[OX1])[O,N]).

[cing]

Love the idea, but I believe that SMARTS pattern matches many more molecules than you intend (you can search it in PubChem and see some of the non-amino acid results), and misses a significant number of non-natural amino acids that contain non-canonical backbones (beta amino acids, for instance, though surely more SMARTS could be made).

Another option would be to piggy-back on NCAA lists already curated like Croitoru et al.. Or, something that we've done, is parse all chEMBL monomers that contain _R1 and _R2 connectivity (as amino acids do) from the chEMBL monomer database (and you get the reproducibility benefit of being able to use a specific chEMBL release).

[jchodera]

@cing: This is a great suggestion!

Do you happen to have a pointer to more documentation about the ChEMBL monomer library? The link is just a download of the library data---is there anything that describes what it contains and how it was constructed, how extensive it is, etc?

Some related tools I found:

• The chembl-downloader seems like a useful tool for pulling the monomer library into RDKit molecules, and easy to use functionality for doing exactly this was added to the most recent release
• There is RDKit functionality for creating simple peptides from these monomers to generate chemical diversity around the amide peptide linker.

Any other suggestions you might have here could be useful.

Eventually, for example, we'd love to work with all major biopolymers, including nucleic acids.

[cing]

I don't have a resource on the composition or how the ChEMBL monomer library was constructed, but it is "stable" in that they've been releasing the monomer library since 2015. ChEMBL team may respond to an email about it. As far as I understand, it's defined entirely by molecules that are deposited in ChEMBL so it likely has gaps (perhaps missing PTMs and no nucleic acids). There's a versioned "HELM Core Monomer" library files here with presumably overlapping content, and RNA building blocks which may be of interest.

I'm not sure about Helm Core, but the ChEMBL XML format is written in the HELM specification here (PDF), and is pretty well documented.

What I can say is that we've had no trouble making free, capping amino acids using molzip functionality described in the blog post above wth ChEMBL monomers. Making arbitrary cyclic peptides through a number of linkages is also doable. I can certainly help on that, but I suppose your use-case is simpler.

[jchodera]

From openmm/openmm#3770 (comment), here's a great illustration of common post-translational modifications (from this article):