next up previous contents index
Next: User-defined energy terms Up: User-defined features and restraint Previous: User-defined feature types   Contents   Index

User-defined restraint forms

To create a new restraint form, derive a new class from the base forms.restraint_form. You should then override the following functions: __init__, eval, vmin, rvmin, min_mean, vheavy, rvheavy, and heavy_mean. Note that presently you can only derive from this base class, not from MODELLER built-in forms.

Restraint forms can act on one or more features (each of which has an accompanying integer modality, which you can use for any purpose), and can take any number of floating-point parameters as input. The features and parameters are stored in self._features and self._parameters respectively, but for convenience the base constructor restraint_form.__init__ can set initial values for these.

The eval function is called from MODELLER with the current feature values, their types and modalities, and the parameter vector. You should return the objective function contribution and, if requested, the derivatives with respect to each feature. The feature types are required by the deltaf function, which returns the difference between the current feature value and the mean (a simple subtraction is not sufficient, as some feature types are periodic). Note that you must use the passed parameter vector, as the class is not persistent, and as such the self._parameters variable (or any other object variable you may have set) is not available to this function.

The other functions are used to return the minimal and heavy restraint violations (both absolute and relative; see Section 5.3.1) and the means. The heavy and minimal means correspond to the global and local minima.

Example: examples/python/user_form.py


from modeller import *
from modeller.scripts import complete_pdb

env = environ()

env.io.atom_files_directory = '../atom_files'
log.verbose()
env.libs.topology.read(file='$(LIB)/top_heav.lib')
env.libs.parameters.read(file='$(LIB)/par.lib')

class mygauss(forms.restraint_form):
    """An implementation of Modeller's harmonic/Gaussian restraint (type 3)
       in pure Python"""

    rt = 0.5900991    # RT at 297.15K, in kcal/mol

    def __init__(self, group, feature, mean, stdev):
        forms.restraint_form.__init__(self, group, feature, 0, (mean, stdev))

    def eval(self, feats, iftyp, modal, param, deriv):
        (mean, stdev) = param
        delt = self.deltaf(feats[0], mean, iftyp[0])
        val = self.rt * 0.5 * delt**2  / stdev**2
        if deriv:
            fderv = self.rt * delt / stdev**2
            return val, [fderv]
        else:
            return val

    def vmin(self, feats, iftyp, modal, param):
        (mean, stdev) = param
        return self.deltaf(feats[0], mean, iftyp[0])

    def rvmin(self, feats, iftyp, modal, param):
        (mean, stdev) = param
        return self.deltaf(feats[0], mean, iftyp[0]) / stdev

    def min_mean(self, feats, iftyp, modal, param):
        (mean, stdev) = param
        return [mean]

    # There is only one minimum, so the 'heavy' mean is the same as the 'min'
    vheavy = vmin
    rvheavy = rvmin
    heavy_mean = min_mean

mdl = complete_pdb(env, "1fdx")
sel = selection(mdl)
rsr = mdl.restraints
at = mdl.atoms
rsr.add(mygauss(group=physical.bond,
                feature=features.distance(at['CB:1'], at['CA:1']),
                mean=1.5380, stdev=0.0364))
sel.energy()


next up previous contents index
Next: User-defined energy terms Up: User-defined features and restraint Previous: User-defined feature types   Contents   Index
Ben Webb 2007-08-03