radifar / deemian Goto Github PK

Interaction Pseudo-Atoms (IPA) (Desaphy et al, 2013) seems to be coveted by many medicinal chemist, it will be even better if Deemian also capable of aligning several macromolecule so that the IPA can be used to identifiy pharmacophore points or hotspots.

In the future there will be (at least) 3 kind of output:

• interactions (containing info about the atom pair, distance, and interaction type)
• bitstrings
• IPA (Interaction Pseudo-Atom) (Desaphy et al 2013)

The metadata should also include the output type for each of its parquet file manifest data.

Flake8 by default only allow 80 as max line length, which is inconsistent with Black setting (119).

Create .flake8 setup file and set max line length to 119 to make it consistent with Black.

Part of version 0.1 #1

Using getaligntransform method of RDKit, it is possible to align macromolecule based on selection. The basic idea is to create selection on several PDB molecule, then use the reference selection as the target for alignment.

The basic syntax is generally like this (subject to change):

molecule 5nzn.pdb [
	select protein_A_ref = chain A and protein
	select active_site_ref_1 = chain A and resid 116 to 120
	select active_site_ref_2 = chain A and resid 150 to 155
	select active_site_ref_3 = chain A and resid 225 to 250
	select active_site_ref_4 = chain A and resid 291 to 295
	select active_site_ref_5 = chain A and resid 366 to 370
	combine_selection active_site_ref = active_site_ref_1, active_site_ref_2, active_site_ref_3, active_site_ref_4, active_site_ref_5
	select oseltamivir = chain A and resname G39
	assign bond oseltamivir template CCC(CC)O[C@@H]1C=C(C[C@@H]([C@H]1NC(=O)C)N)C(=O)O
]

molecule 1f8b.pdb [
	select protein_A_dan = chain A and protein
	select active_site_1 = chain A and resid 116 to 120
	select active_site_2 = chain A and resid 150 to 155
	select active_site_3 = chain A and resid 225 to 250
	select active_site_4 = chain A and resid 291 to 295
	select active_site_5 = chain A and resid 366 to 370
	combine_selection active_site = active_site_1, active_site_2, active_site_3, active_site_4, active_site_5
	select daneuraminic = chain A and resname DAN

	align active_site to active_site_ref
]

measure ipa_combined [
	ipa subject_1, center, subject_2
	ionizable positive true
	ionizable negative true
	between oseltamivir and protein_A
    conformation 1
]

What happen in the script above is to create a series of selection based on area around active site. The C-alpha coordinates for both active site is then extracted and transformation matrix is retrieved using getaligntransform method of RDKit. After that this transformation matrix is used to transform the whole structure of 1f8b model. This approach is better because it allows a more selective transformation rather than aligning the whole structure, as this selective transformation allows more aligned IPA generation.

Read PDB file and use its data to create Pandas DataFrame. Which can be very useful for creating selections and identifying interactions (using KDTree).