Bio Structural Biology Structure Io

Version Compatibility

Reference examples tested with: biopython 1.85+

Before using code patterns, verify installed versions match. If versions differ:

• Python: pip show <package> then help(module.function) to check signatures

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Structure I/O

"Read a structure file" -> Parse a deposited coordinate file into an in-memory SMCRA tree, or fetch it from a wwPDB mirror.

• Python: Bio.PDB.MMCIFParser().get_structure('id', 'file.cif'), Bio.PDB.PDBParser(), Bio.PDB.PDBList()

Governing Principle

mmCIF (PDBx) is the canonical modern format; the legacy fixed-column PDB format is a frozen, lossy container, and a "parse everything as PDB" reflex silently truncates or fails on anything large. The PDB format was frozen in 2012 and cannot physically exceed 99,999 atoms (5-digit serial), ~62 chains (single alphanumeric chain id), or 9,999 residues per chain (wwPDB file-format documentation; the PDB archive itself is Berman et al. 2000 Nucleic Acids Res 28:235-242). Large assemblies (ribosomes, capsids, spliceosomes, most big cryo-EM structures) therefore exist ONLY as mmCIF, and converting a big mmCIF down to PDB renames multi-character chains and overflows serial numbers, silently corrupting any downstream tool that keys on chain id. mmCIF has been the wwPDB archive standard since 2014 and mandatory for crystallographic depositions since July 2019.

Two further traps compound this. First, Bio.PDB is PERMISSIVE by design: it reads malformed files, and it silently drops anisotropic B-factors (ANISOU), collapses each disordered atom to its highest-occupancy alternate, and never models most metadata (resolution, method, R-free, entity graph, assembly operators). Parse-success is not data integrity. Second, the deposited coordinates for an X-ray entry are usually the ASYMMETRIC UNIT, a crystallographic bookkeeping object that is frequently NOT the biologically functional oligomer -- so any interface, oligomeric-state, or buried-surface question must first obtain the biological assembly (Krissinel & Henrick 2007 J Mol Biol 372:774). "One chain in the file" is never evidence of a monomer.

The escape hatch for all three ceilings (assembly generation, very large structures, full mmCIF fidelity) is gemmi (Wojdyr 2022 JOSS 7:4200); Bio.PDB cannot apply the assembly operators itself. Prefer Bio.PDB for teaching, small structures, and hierarchy walks; reach for gemmi when the questions above appear.

Decision: which format

Decision: Bio.PDB vs gemmi

Required Imports

from Bio.PDB import PDBParser, MMCIFParser, PDBIO, MMCIFIO, PDBList, Select
from Bio.PDB.MMCIF2Dict import MMCIF2Dict
from Bio.PDB.binary_cif import BinaryCIFParser

Parse an mmCIF File (the modern default)

from Bio.PDB import MMCIFParser

# auth_residues/auth_chains default to True: numbering matches the paper/UniProt.
parser = MMCIFParser(QUIET=True)
structure = parser.get_structure('4hhb', '4hhb.cif')

# label numbering is contiguous 1..N with no insertion codes -- a DIFFERENT scheme.
label_parser = MMCIFParser(QUIET=True, auth_residues=False, auth_chains=False)
label_structure = label_parser.get_structure('4hhb', '4hhb.cif')

Setting auth_residues=False renumbers to the mmCIF internal label scheme, so residue 100 in one parse is a different residue in the other. This is the single most common "my selection points at the wrong residue" bug. auth is what matches the literature and sequence databases; label is gap-free internal bookkeeping. Pick one scheme and stay in it.

Parse a Legacy PDB File

from Bio.PDB import PDBParser

# QUIET=True suppresses PDBConstructionWarning (discontinuous chains, missing occupancy).
parser = PDBParser(QUIET=True)
structure = parser.get_structure('1crn', '1crn.pdb')

Parse a BinaryCIF File (compact binary, replaces MMTF)

from Bio.PDB.binary_cif import BinaryCIFParser

# .get_structure(id, source); gz is handled transparently.
parser = BinaryCIFParser()
structure = parser.get_structure('1gbt', '1gbt.bcif.gz')

MMTF is intentionally absent here. RCSB retired it on 2 July 2024 and MMTFParser.get_structure_from_url targets a decommissioned service; BinaryCIF is the replacement.

Read Metadata Bio.PDB Drops (MMCIF2Dict)

from Bio.PDB.MMCIF2Dict import MMCIF2Dict

# MMCIF2Dict returns category -> list[str]; index [0] and cast yourself.
meta = MMCIF2Dict('4hhb.cif')
resolution = meta.get('_refine.ls_d_res_high', ['NA'])[0]
method = meta.get('_exptl.method', ['NA'])[0]
r_free = meta.get('_refine.ls_R_factor_R_free', ['NA'])[0]
r_work = meta.get('_refine.ls_R_factor_R_work', ['NA'])[0]

The parser's thin structure.header omits resolution/R-free for many files; the dict reaches anything in the mmCIF, including assembly operators the object model never builds.

Download from RCSB (PDBList)

from Bio.PDB import PDBList

pdbl = PDBList()

# pdir=None writes into a two-char divided subdirectory tree (e.g. hh/4hhb.cif),
# NOT the current directory; pass pdir='.' to control the location.
path = pdbl.retrieve_pdb_file('4HHB', pdir='.', file_format='mmCif')

# file_format 'pdb' fetches legacy PDB only when the entry fits the format.
legacy_path = pdbl.retrieve_pdb_file('4HHB', pdir='.', file_format='pdb')

file_format='mmCif' (that exact casing) is the current default recommendation. retrieve_pdb_file has no assembly_num parameter in current Biopython -- download the assembly directly (below).

Download the Biological Assembly (Bio.PDB cannot build it)

import gzip, shutil, urllib.request

# The ASU is often not the functional oligomer; RCSB pre-applies the operators
# in the -assemblyN file, so downloading it is safer than regenerating.
pdb_id = '1abc'
url = f'https://files.rcsb.org/download/{pdb_id.upper()}-assembly1.cif.gz'
urllib.request.urlretrieve(url, f'{pdb_id}-assembly1.cif.gz')
with gzip.open(f'{pdb_id}-assembly1.cif.gz', 'rb') as fin, open(f'{pdb_id}-assembly1.cif', 'wb') as fout:
    shutil.copyfileobj(fin, fout)

Bio.PDB has no operator-application code, so if only the deposited ASU is on disk it cannot construct the assembly -- use the RCSB assembly file or gemmi's transform_to_assembly.

Write a Structure

from Bio.PDB import PDBParser, MMCIFIO

parser = PDBParser(QUIET=True)
structure = parser.get_structure('1crn', '1crn.pdb')

# Writing mmCIF preserves multi-char chains and >99,999 serials; PDB cannot.
io = MMCIFIO()
io.set_structure(structure)
io.save('1crn_out.cif')

Write a Subset with the Select Class

from Bio.PDB import PDBParser, PDBIO, Select

class ProteinChainSelect(Select):
    def __init__(self, chain_id):
        self.chain_id = chain_id

    def accept_chain(self, chain):
        return chain.id == self.chain_id

    def accept_residue(self, residue):
        # id[0] is the hetflag: ' ' standard, 'W' water, 'H_XXX' hetero.
        return residue.id[0] == ' '

parser = PDBParser(QUIET=True)
structure = parser.get_structure('1crn', '1crn.pdb')

io = PDBIO()
io.set_structure(structure)
io.save('chain_A_protein.pdb', ProteinChainSelect('A'))

Override any of accept_model, accept_chain, accept_residue, accept_atom to return truthy to keep. Writing a large mmCIF back out as PDB through PDBIO is where multi-character chains and serial overflow silently corrupt the output.

Capture Warnings for an Unfamiliar File

from Bio.PDB import PDBParser
import warnings

# QUIET=True is the reflex, but it hides 'chain is discontinuous' -- the warning
# that flags a numbering gap or a merge the parser should not have made.
parser = PDBParser(QUIET=False)
with warnings.catch_warnings(record=True) as caught:
    warnings.simplefilter('always')
    structure = parser.get_structure('unknown', 'unknown.pdb')
    for w in caught:
        print(w.message)