Accessible Volume Restraint

Accessible volume (AV) restraints can be used for integrative modeling and for scoring structures. There are two versions of AV restraints IMP.bff.AVNetworkRestraint and IMP.bff.restraints.AVNetworkRestraintWrapper. The AV restraint wrapped for PMI, AVNetworkRestraintWrapper provides a mean dye positions, an approximation that is faster when working with rigid bodies.

In this example a T4 lysozme structure is loaded and scored using IMP.bff.AVNetworkRestraint by experimental data in a fps.json file.

First import all necessary libraries.

import pathlib
import json

import numpy as np
import pylab as plt

import RMF
import IMP
import IMP.rmf
import IMP.atom
import IMP.bff

Create a model and populate a IMP.Hierarchy.

m = IMP.Model()
pdb_fn = pathlib.Path(IMP.bff.get_example_path('structure')) / "T4L/3GUN.pdb"
hier = IMP.atom.read_pdb(str(pdb_fn), m)

The AV network restraint takes a fps.json file as an input. fps.json files can contain multiple sets of distances for scoring structures.

fps_json_path = IMP.bff.get_example_path("structure/T4L/fret.fps.json")
with open(fps_json_path) as fp:
    fps_json = json.load(fp)

score_set_c1 = "chi2_C2_33p"
fret_restraint = IMP.bff.AVNetworkRestraint(hier, str(fps_json_path), score_set=score_set_c1)

The score is computed

v = fret_restraint.unprotected_evaluate(None)
print("Score: {:.1f}".format(v))

Score: 12.6

The computed model distances and the experimental distances contained IMP.bff.AVNetworkRestraint objects can be accesses as follows, e.g., to identify outliers.

experimental_distances = fret_restraint.get_used_distances()

model = list()
experiment = list()
for key in experimental_distances:
    e_dist = experimental_distances[key]
    d = json.loads(e_dist.get_json())
    model_distance = fret_restraint.get_model_distance(
        d["position1_name"],
        d["position2_name"],
        d["Forster_radius"],
        d["distance_type"]
    )
    e_dist.score_model(model_distance)
    model.append(model_distance)
    experiment.append(d['distance'])

Plot the experiment against the model distance

x = np.linspace(25, 60)
plt.plot(model, experiment, "o")
plt.plot(x, x, "-")
plt.xlabel("Model distance [Ang.]")
plt.ylabel("Experimental distance [Ang.]")
plt.show()