Add the ability for the PDBManager to perform interface-based chain filtering
#333
Conversation
graphein/ml/datasets/pdb_data.py
Outdated
| @@ -23,6 +25,11 @@ | |||
| ) | |||
| from graphein.utils.dependencies import is_tool | |||
|
|
|||
| PRIMARY_INTERCHAIN_CONTACT_ATOMS_FOR_FILTERING: List[str] = ["CA", "C4'"] | |||
| SECONDARY_INTERCHAIN_CONTACT_ATOMS_NOT_FOR_FILTERING: List[str] = ["H"] | |||
| PRIMARY_HYDROGEN_BOND_ATOMS_FOR_FILTERING: List[str] = ["N", "O", "N1", "N9", "N3", "C2", "C4", "C5", "C6"] | |||
There was a problem hiding this comment.
This should be vetted more carefully, as I initially chose these atom types heuristically.
There was a problem hiding this comment.
What is this atom naming scheme? It doesn't ring any bells for me (
graphein/graphein/protein/resi_atoms.py
Line 276 in 281ce30
There was a problem hiding this comment.
We already have these constants:
graphein/graphein/protein/resi_atoms.py
Line 858 in 281ce30
graphein/graphein/protein/resi_atoms.py
Line 880 in 281ce30
There was a problem hiding this comment.
What is this atom naming scheme? It doesn't ring any bells for me (
graphein/graphein/protein/resi_atoms.py
Line 276 in 281ce30
)
The N, CA, O, and H atoms correspond to regular protein vocabulary, however, all other types correspond to nucleic acid residue atoms. My initial goal with this PR was to make a generic dataset chain filter for protein-protein interactions, protein-nucleic acid interactions, and nucleic acid-nucleic acid interactions (inspired by the dataset curation technique of RoseTTAFold2NA for protein-nucleic acid structure prediction - https://www.biorxiv.org/content/10.1101/2022.09.09.507333v1.full.pdf - page 8). I am essentially trying to reproduce this filtering logic with the PDBManager (minus all the sequence alignments), and I thought a PR would be in order.
There was a problem hiding this comment.
Per a suggestion from a colleague, I have removed the C atoms from the hydrogen bond calculation, as these atoms are very rarely involved in the formation of h-bonds in proteins and NAs.
There was a problem hiding this comment.
The N, CA, O, and H atoms correspond to regular protein vocabulary, however, all other types correspond to nucleic acid residue atoms.
Got it, bells ring for me now :)
So these H-bond definitions do not account for sidechain-X hbonds, only backbone-backbone hbonds?
There was a problem hiding this comment.
Right. Here's a naive question on my part: How frequent would you say the occurrence of sidechain-X hbonds is? If they are pretty common, perhaps we can simply include more protein and nucleic acid (NA) atom types to the list here?
There was a problem hiding this comment.
Seemingly quite common!
There was a problem hiding this comment.
By way of how I have designed this filtering logic, I am assuming that each (protein or NA) residue (potentially) contains the following atoms: "N", "O", "N1", "N9", "N3". Given the prevalence of sidechain hbonds, what types of protein atoms (shared across all residue types) would you say would be most reasonable to include to cover most of the possible hbonds mentioned in this article? The only other atom type I think we could include would be the carbon-beta (Cb) atoms.
for more information, see https://pre-commit.ci
|
Kudos, SonarCloud Quality Gate passed!
|
What does this implement/fix? Explain your changes
This allows one to select PDB complex chains satisfying certain interface contact or hydrogen bonding constraints.