Homology Modeling

Template-based protein structure prediction.

MODELLER

from modeller import *
from modeller.automodel import *

# Set up environment
env = Environ()
env.io.atom_files_directory = ['.', 'atom_files']

# Create alignment
aln = Alignment(env)
aln.append(file='alignment.ali')

# Build model
a = AutoModel(env, alnfile='alignment.ali',
              knowns='template', sequence='target')
a.starting_model = 1
a.ending_model = 5
a.make()

Template Selection

from Bio.Blast import NCBIWWW, NCBIXML

def find_templates(sequence, evalue_threshold=0.001):
    """Find structural templates via BLAST."""
    result = NCBIWWW.qblast("blastp", "pdb", sequence)
    records = NCBIXML.parse(result)

    templates = []
    for record in records:
        for alignment in record.alignments:
            for hsp in alignment.hsps:
                if hsp.expect < evalue_threshold:
                    templates.append({
                        'pdb_id': alignment.hit_id.split('|')[1],
                        'evalue': hsp.expect,
                        'identity': hsp.identities / hsp.align_length
                    })
    return templates

Model Quality Assessment

# DOPE score in MODELLER
def assess_model(model_file):
    env = Environ()
    mdl = Model(env, file=model_file)
    atmsel = Selection(mdl)
    score = atmsel.assess_dope()
    return score