filter molecule for templates

[fgrunewald]

This PR sets the stage for some important changes:

• instead of matching templates by resnames each residue now requires a template_name attribute. This allows matching of non-identical residues that has the same resname
• extract blocks now takes a template graph as input to extract the interactions, which is also used in PR Convert itp files to ff files #327

To Do

• move ExtractBlock to separate file something like manipulate_molecules.py
• speed up equivalence check or at least enable a bypass

@pckroon do you have any idea how to do the code below in faster?

polyply_1.0/polyply/src/generate_templates.py

Lines 295 to 314 in b97ba71

	def group_by_isomorphism(meta_molecule, template_graphs={}):
	"""
	Extract all unique fragment graphs from meta_molecule
	using the full subgraph isomorphism check.
	"""
	template_graphs = {}
	for node in meta_molecule.nodes:
	resname = meta_molecule.nodes[node]["resname"]
	graph = meta_molecule.nodes[node]["graph"]
	if resname in template_graphs and not nx.is_isomorphic(graph,
	template_graphs[resname],
	node_match=_atoms_match,
	):
	template_name = resname + str(len(template_graphs))
	meta_molecule.nodes[node]["template"] = template_name
	template_graphs[template_name] = graph
	else:
	meta_molecule.nodes[node]["template"] = resname
	template_graphs[resname] = graph
	return template_graphs

[fgrunewald]

note to self vf2pp from networkx 3 might speed up matching process also proper line by line benchmark is needed

[pckroon]

Not fundamentally, no. I'm also not sure what the goal you're trying to achieve is.

[fgrunewald]

Currently polyply assumes all residues are unique. That is mostly true but not quite at least at the all-atom level. For example, there might be a hydrogen extra or less depending on if a residue is located in the center or at the terminal. This functionality checks each residue against a collection of unique residues in order to find those that have the same resname but deviate as graphs. It makes generating the templates more robust for complicated AA structures

[pckroon]

Alright. I don't think there's really a faster way then. Unless you are willing to accept approximation, e.g. only check atom names.