score.py

#!/usr/bin/env python

__doc__ = "Score each of a set of combinations."

# analyse the ensemble, first we will do the rmsd stuff
import IMP.multifit
from IMP import OptionParser


def get_color_map():
    colors = {}
    colors["Rpt1"] = [0.78, 0.78, 0.73]
    colors["Rpt2"] = [0.78, 0.66, 0.58]
    colors["Rpt3"] = [0.77, 0.43, 0.5]
    colors["Rpt4"] = [0.76, 0.29, 0.67]
    colors["Rpt5"] = [0.51, 0.14, 0.75]
    colors["Rpt6"] = [0.0, 0., 0.75]
    colors["Rpn1"] = [0.34, 0.36, 0.27]
    colors["Rpn2"] = [0.42, 0.43, 0.36]
    colors["Rpn3"] = [0.49, 0.5, 0.44]
    colors["Rpn5"] = [0.56, 0.57, 0.51]
    colors["Rpn6"] = [0.64, 0.64, 0.59]
    colors["Rpn7"] = [0.71, 0.71, 0.66]
    colors["Rpn8"] = [0.78, 0.78, 0.74]
    colors["Rpn9"] = [1, 0, 0]
    colors["Rpn10"] = [0, 1, 0]
    colors["Rpn11"] = [0, 0, 1]
    colors["Rpn12"] = [0.5, 0.2, 0.4]
    colors["a1"] = [0.78, 0.78, 0.73]
    colors["a2"] = [0.78, 0.66, 0.58]
    colors["a3"] = [0.77, 0.43, 0.5]
    colors["a4"] = [0.76, 0.29, 0.67]
    colors["a5"] = [0.51, 0.14, 0.75]
    colors["a6"] = [0.0, 0., 0.75]
    colors["a7"] = [0.34, 0.36, 0.27]
    colors["a8"] = [0.42, 0.43, 0.36]
    colors["a9"] = [0.49, 0.5, 0.44]
    colors["a10"] = [0.56, 0.57, 0.51]

    colors["a11"] = [0.78, 0.78, 0.73]
    colors["a12"] = [0.78, 0.66, 0.58]
    colors["a13"] = [0.77, 0.43, 0.5]
    colors["a14"] = [0.76, 0.29, 0.67]
    colors["a15"] = [0.51, 0.14, 0.75]
    colors["a16"] = [0.0, 0., 0.75]
    colors["a17"] = [0.34, 0.36, 0.27]
    colors["a18"] = [0.42, 0.43, 0.36]
    colors["a19"] = [0.49, 0.5, 0.44]
    colors["a20"] = [0.56, 0.57, 0.51]

    colors["a21"] = [0.78, 0.78, 0.73]
    colors["a22"] = [0.78, 0.66, 0.58]
    colors["a23"] = [0.77, 0.43, 0.5]
    colors["a24"] = [0.76, 0.29, 0.67]
    colors["a25"] = [0.51, 0.14, 0.75]
    colors["a26"] = [0.0, 0., 0.75]
    colors["a27"] = [0.34, 0.36, 0.27]
    colors["a28"] = [0.42, 0.43, 0.36]
    colors["a29"] = [0.49, 0.5, 0.44]
    colors["a30"] = [0.56, 0.57, 0.51]
    return colors


def decompose(dmap, mhs):
    full_sampled_map = IMP.em.SampledDensityMap(dmap.get_header())
    all_ps = []
    for mh in mhs:
        all_ps += IMP.core.get_leaves(mh)
    full_sampled_map.set_particles(all_ps)
    full_sampled_map.resample()
    full_sampled_map.calcRMS()
    upper = (
        dmap.get_number_of_voxels(
        ) * dmap.get_header(
        ).dmean * full_sampled_map.get_header(
        ).dmean) / len(
        mhs)
    lower = dmap.get_number_of_voxels(
    ) * dmap.get_header(
    ).rms * full_sampled_map.get_header(
    ).rms
    norm_factors = [upper, lower]
    print "===============my norm factors:", upper, lower
    return norm_factors


def score_each_protein(dmap, mhs, sd):
    norm_factors = decompose(dmap, mhs)
    scores = []
    cc = IMP.em.CoarseCC()
    mdl = mhs[0].get_model()
    for i in range(len(mhs)):
        leaves = IMP.core.get_leaves(mhs[i])
        rb = IMP.core.RigidMember(leaves[0]).get_rigid_body()
        mh_dmap = IMP.em.SampledDensityMap(dmap.get_header())
        mh_dmap.set_particles(leaves)
        mh_dmap.resample()
        mh_dmap.calcRMS()
        fits = IMP.multifit.read_fitting_solutions(
            sd.get_component_header(i).get_transformations_fn())
        mh_scores = []
        for fit in fits[:15]:
            IMP.core.transform(rb, fit.get_fit_transformation())
            mh_dmap.resample()
            mh_scores.append(
                cc.cross_correlation_coefficient(
                    dmap,
                    mh_dmap,
                    0.,
                    False,
                    norm_factors))
            IMP.core.transform(rb, fit.get_fit_transformation().get_inverse())
        scores.append(mh_scores)
        print "=====mol", i, mh_scores
    return scores


def usage():
    usage = """%prog [options] <asmb> <asmb.proteomics> <asmb.mapping>
           <alignment.params> <combinatins> <score combinations [output]>

Score each of a set of combinations.
"""
    parser = OptionParser(usage)
    parser.add_option("-m", "--max", dest="max", default=999999999,
                      help="maximum number of fits considered")
    (options, args) = parser.parse_args()
    if len(args) != 6:
        parser.error("incorrect number of arguments")
    return [options, args]


def run(asmb_fn, proteomics_fn, mapping_fn, params_fn, combs_fn,
        scored_comb_output_fn, max_comb):
    asmb = IMP.multifit.read_settings(asmb_fn)
    dmap = IMP.em.read_map(asmb.get_assembly_header().get_dens_fn())
    dmap.get_header().set_resolution(
        asmb.get_assembly_header().get_resolution())
    dmap.update_voxel_size(asmb.get_assembly_header().get_spacing())
    dmap.set_origin(asmb.get_assembly_header().get_origin())
    threshold = asmb.get_assembly_header().get_threshold()
    combs = IMP.multifit.read_paths(combs_fn)
    # get rmsd for subunits
    colors = get_color_map()
    names = colors.keys()
    print params_fn
    alignment_params = IMP.multifit.AlignmentParams(params_fn)
    alignment_params.show()

    IMP.base.set_log_level(IMP.base.TERSE)
    print "=========", combs_fn
    combs = IMP.multifit.read_paths(combs_fn)
    print "=========1"
    # sd=IMP.multifit.read_settings(asmb_fn)
    print "=========2"
    prot_data = IMP.multifit.read_proteomics_data(proteomics_fn)
    print "=========3"
    mapping_data = IMP.multifit.read_protein_anchors_mapping(
        prot_data, mapping_fn)
    print "=========4"
    em_anchors = mapping_data.get_anchors()
    print "=========5"
    ensmb = IMP.multifit.Ensemble(asmb, mapping_data)
    print "=========6"
    # load all proteomics restraints
    align = IMP.multifit.ProteomicsEMAlignmentAtomic(
        mapping_data, asmb, alignment_params)
    align.set_fast_scoring(False)
    print "align"
    mdl = align.get_model()
    mhs = align.get_molecules()
    align.add_states_and_filters()
    rbs = align.get_rigid_bodies()
    print IMP.core.RigidMember(IMP.core.get_leaves(mhs[0])[0]).get_rigid_body()
    align.set_density_map(dmap, threshold)
    gs = []
    for i, mh in enumerate(mhs):
        ensmb.add_component_and_fits(mh,
                                     IMP.multifit.read_fitting_solutions(asmb.get_component_header(i).get_transformations_fn()))
        try:
            rgb = colors[mh.get_name()]
        except:
            rgb = colors[names[i]]
        color = IMP.display.Color(rgb[0], rgb[1], rgb[2])
        '''
        for p in IMP.core.get_leaves(mh):
            g= IMP.display.XYZRGeometry(p)
            g.set_color(color)
            gs.append(g)
        '''
    all_leaves = []
    for mh in mhs:
        all_leaves += IMP.core.XYZs(IMP.core.get_leaves(mh))

    align.add_all_restraints()
    print "====1"
    print "Get number of restraints:", len(mdl.get_restraints())
    rs = mdl.get_restraints()
    for r in mdl.get_restraints():
        rr = IMP.kernel.RestraintSet.get_from(r)
        for i in range(rr.get_number_of_restraints()):
            print rr.get_restraint(i).get_name()
    output = open(scored_comb_output_fn, "w")
    # load ref structure
    ref_mhs = []
    all_ref_leaves = []
    for i in range(asmb.get_number_of_component_headers()):
        c = asmb.get_component_header(i)
        fn = c.get_reference_fn()
        if fn:
            ref_mhs.append(IMP.atom.read_pdb(fn, mdl))
            all_ref_leaves += IMP.core.get_leaves(ref_mhs[-1])
    for r in rs:
        rr = IMP.kernel.RestraintSet.get_from(r)
        for i in range(rr.get_number_of_restraints()):
            output.write(rr.get_restraint(i).get_name() + "|")
    output.write("\n")
    # add fit restraint
    fitr = IMP.em.FitRestraint(all_leaves, dmap)
    mdl.add_restraint(fitr)
    print "Number of combinations:", len(combs[:max_comb])

    print "native score"
    num_violated = 0
    for r in rs:
        rr = IMP.kernel.RestraintSet.get_from(r)
        for j in range(rr.get_number_of_restraints()):
            print rr.get_restraint(j).get_name(), rr.evaluate(False)

    prev_name = ''
    for i, comb in enumerate(combs[:max_comb]):
        print "Scoring combination:", comb
        ensmb.load_combination(comb)
        num_violated = 0
        for r in rs:
            rr = IMP.kernel.RestraintSet.get_from(r)
            for j in range(rr.get_number_of_restraints()):
                print rr.get_restraint(j).get_name()
                rscore = rr.evaluate(False)
                if rscore > 5:
                    num_violated = num_violated + 1
        IMP.atom.write_pdb(mhs, "model.%d.pdb" % (i))
        print str(all_leaves[0]) + " :: " + str(all_leaves[-1])
        score = mdl.evaluate(False)
        num_violated = 0
        msg = "COMB" + str(i) + "|"
        for r in rs:
            rr = IMP.kernel.RestraintSet.get_from(r)
            for j in range(rr.get_number_of_restraints()):
                current_name = rr.get_restraint(j).get_name()
                if current_name != prev_name:
                    msg += ' ' + current_name + ' '
                    prev_name = current_name
                rscore = rr.get_restraint(j).evaluate(False)
                msg += str(rscore) + "|"
                if rscore > 5:
                    num_violated = num_violated + 1
        # msg+="|"+str(score)+"|"+str(num_violated)+"|\n"
        msg += "|" + str(
            score) + "|" + str(
            num_violated) + "||||" + str(
            fitr.evaluate(False)) + "||:"
        if all_ref_leaves:
            msg += str(IMP.atom.get_rmsd(IMP.core.XYZs(all_leaves),
                                         IMP.core.XYZs(all_ref_leaves)))
        output.write(msg + "\n")
        print msg
        ensmb.unload_combination(comb)
    output.close()


def main():
    (options, args) = usage()
    run(args[0], args[1], args[2], args[3], args[4], args[5], int(options.max))

if __name__ == "__main__":
    main()