PDB to mmCIF

[glevans]

Hi Marcin,

I've been working (among other things) on a project as consultant with beamline scientists and developers who have a pipeline where due to legacy code they can only produce PDB files.

I made a small script that fixes missing TER records to the PDB file.

import argparse

def add_ter(input_file, output_file):
    """Reads a PDB file, inserts a TER record between ATOM and HETATM records,
    and writes the result to a new file.

    Args:
        input_file (str): Path to the input PDB file.
        output_file (str): Path to the output PDB file.
    """
    with open(input_file, 'r') as infile, open(output_file, 'w') as outfile:
        previous_line_was_atom = False
        for line in infile:
            is_atom = line.startswith('ATOM')
            is_hetatm = line.startswith('HETATM')

            if is_hetatm and previous_line_was_atom:
                outfile.write('TER\n')

            outfile.write(line)

            previous_line_was_atom = is_atom

if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='Add TER record between ATOM and HETATM records in a PDB file.')
    parser.add_argument('input_file', help='The input PDB file.')
    parser.add_argument('output_file', help='The output PDB file.')
    args = parser.parse_args()

    add_ter(args.input_file, args.output_file)
    print(f'Processed {args.input_file} and saved the output to {args.output_file}')

I can see that we can use Gemmi to generate the correct mmCIF.
I could test this using your webtools available here:
https://project-gemmi.github.io/wasm/convert/pdb2cif.html

I forked from the following repo (and have now deleted):
https://github.com/project-gemmi/wasm

I thought maybe I could figure out what the backend for the webtools were
--> if it was Python anyway...
However, as you know the back-end is C++ not Python.

I am sure there is correct set of steps with the Python Gemmi library
that one could use to do the same thing as the webtool PDB2CIF.
Unfortunately I have not figured it out yet.

Any chance you or one of your colleagues could help?

[glevans]

BTW to goal if this project is to generate files for wwPDB deposition.

[MartinMalyMM]

Hi Genieve!
Isn't gemmi convert file.pdb file.mmcif what you are looking for? All options can be listed using gemmi convert -h .

[wojdyr]

All refinement programs can generate also mmCIF file nowadays, so ideally the beamline developers would update their pipeline.

[wojdyr]

And in Python its:

import gemmi

def convert_pdb_to_mmcif(input_pdb, output_cif):
    # 1. Read the PDB file
    # The read_structure function automatically detects the format
    st = gemmi.read_structure(input_pdb)
    
    # 2. Setup the mmCIF document
    # make_mmcif_document() converts the internal structure object 
    # into a CIF document object
    doc = st.make_mmcif_document()
    
    # 3. Write the file
    doc.write_file(output_cif)
    print(f"Successfully converted {input_pdb} to {output_cif}")

# Usage
convert_pdb_to_mmcif("example.pdb", "example.cif")

[glevans]

Thanks so much for the rapidly provided insight :-)

To capture here:
The PDB file from this pipeline was generated by REFMAC 5.8.0267
The TER records needed to first be removed (wrong placement) using script below.
Then added in the right place using script above.

Here is the small script I used the strip TER records from the PDB file.
I think it likely there is a way to strip TER records in Gemmi too.

import argparse

def remove_ter(input_file, output_file):
   """Reads a file, removes lines containing 'TER', and writes the result to a new file.

   Args:
       input_file (str): Path to the input file.
       output_file (str): Path to the output file.
   """
   with open(input_file, 'r') as infile, open(output_file, 'w') as outfile:
       for line in infile:
           if 'TER' not in line:
               outfile.write(line)

if __name__ == '__main__':
   parser = argparse.ArgumentParser(description='Remove lines containing TER from a file.')
   parser.add_argument('input_file', help='The input file.')
   parser.add_argument('output_file', help='The output file.')
   args = parser.parse_args()

   remove_ter(args.input_file, args.output_file)
   print(f'Processed {args.input_file} and saved the output to {args.output_file}')

Someone in the team at the synchrotron is looking into updating to code so PDB files are no longer used.
However, we needed legacy compatibility.

Thanks again!!

---

BTW gemmi convert gave a nice message complaining about the TER records until I ran the remove_ter.py followed by the add_ter.py

The message:

WARNING: TER records in the input PDB are clearly where they shouldn't be.
WARNING: Ignoring all TER records.

[wojdyr]

After updating CCP4, Refmac version will be:
refmac5; version 5.8.0431
The pipeline should also replace refmac5with refmacat
These two steps should ensure that proper mmCIF is generated.

[CV-GPhL]

Dear Genevieve, remember that the TER record might not only come after ATOM|HETATM but also ANISOU: the documentation [1] isn't explicit about this, but since ANISOU is always related to a ATOM/HETATM record it needs to be taken as part of such a record. Cheers Clemens [1] http://www.wwpdb.org/documentation/file-format-content/format33/sect9.html#ANISOU

…

On Thu, Jan 22, 2026 at 03:49:41PM -0800, Genevieve Evans wrote: Thanks so much for the rapidly provided insight :-) To capture here: The PDB file from this pipeline was generated by REFMAC 5.8.0267 The TER records need to first be removed (wrong placement). Then added in the right place using script above. Here is the small script I used the strip TER records from code. I think it likely there is a way to strip TER records in Gemmi too. ``` import argparse def remove_ter(input_file, output_file): """Reads a file, removes lines containing 'TER', and writes the result to a new file. Args: input_file (str): Path to the input file. output_file (str): Path to the output file. """ with open(input_file, 'r') as infile, open(output_file, 'w') as outfile: for line in infile: if 'TER' not in line: outfile.write(line) if __name__ == '__main__': parser = argparse.ArgumentParser(description='Remove lines containing TER from a file.') parser.add_argument('input_file', help='The input file.') parser.add_argument('output_file', help='The output file.') args = parser.parse_args() remove_ter(args.input_file, args.output_file) print(f'Processed {args.input_file} and saved the output to {args.output_file}') ``` Someone in the team is looking into updating to code so PDB files are no longer used. However, we need legacy compatibility. Thanks again!! -- Reply to this email directly or view it on GitHub: #400 (comment) You are receiving this because you are subscribed to this thread. Message ID: ***@***.***>

-- *-------------------------------------------------------------- * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com * Global Phasing Ltd., 9 Journey Campus, Castle Park * Cambridge CB3 0AX, UK www.globalphasing.com *--------------------------------------------------------------

[MartinMalyMM]

BTW if may help if the beamline pipelines use the latest CCP4 and call the command refmacat instead of refmac5. There was lots of work done by Keitaro and Marcin to improve reading and writing files in last years to sort out issues with strange/incompatible formats.

[glevans]

One of the developers I am working with said I could share the finalized Python function we will be using:

def read_refmac_structure(pdb):
    # this next big adds a missing TER line that is needed for correct conversion to MMCIF
    lines = []
    with open(pdb, 'r') as infile:
        previous_line_was_atom_or_anisou = False
        for line in infile:
            is_atom_or_anisou = line.startswith('ATOM') or line.startswith('ANISOU')
            is_hetatm = line.startswith('HETATM')

            if is_hetatm and previous_line_was_atom_or_anisou:
                lines.append('TER\n')

            if not line.startswith('TER'):
                lines.append(line)

            previous_line_was_atom_or_anisou = is_atom_or_anisou

    # now convert to MMCIF
    struc = gemmi.read_pdb_string(''.join(lines))
    doc = struc.make_mmcif_document()
    scrape_pdb_stats.run(str(pdb), doc)
    return doc

This deals with the ANISOU record complication nicely :-)

Thanks for the insight Clemens!

[wojdyr]

It'd be better to update CCP4 instead.

[CV-GPhL]

Dear all, Not sure I'm reading that code correctly, but it seems as if it assumes that there could not be an ANISOU card after a HETATM one? That's not quite right: even SEQRES records can end with a HETATM record (MSE for example) that has an ANISOU record attached to it ... See "Details" in http://www.wwpdb.org/documentation/file-format-content/format33/sect9.html#TER Cheers Clemens

…

On Fri, Jan 23, 2026 at 02:26:29AM -0800, Genevieve Evans wrote: One of the developers I am working with said I could share the finalized Python function we will be using: ``` def read_refmac_structure(pdb): # this next big adds a missing TER line that is needed for correct conversion to MMCIF lines = [] with open(pdb, 'r') as infile: previous_line_was_atom_or_anisou = False for line in infile: is_atom_or_anisou = line.startswith('ATOM') or line.startswith('ANISOU') is_hetatm = line.startswith('HETATM') if is_hetatm and previous_line_was_atom_or_anisou: lines.append('TER\n') if not line.startswith('TER'): lines.append(line) previous_line_was_atom_or_anisou = is_atom_or_anisou # now convert to MMCIF struc = gemmi.read_pdb_string(''.join(lines)) doc = struc.make_mmcif_document() scrape_pdb_stats.run(str(pdb), doc) return doc ``` This deals with the ANISOU record complication nicely :-) Thanks for the insight Clemens! -- Reply to this email directly or view it on GitHub: #400 (comment) You are receiving this because you commented. Message ID: ***@***.***>

-- *-------------------------------------------------------------- * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com * Global Phasing Ltd., 9 Journey Campus, Castle Park * Cambridge CB3 0AX, UK www.globalphasing.com *--------------------------------------------------------------

[wojdyr]

that code is completely wrong anyway. It puts TER after last ATOM,which is wrong: atoms in polymer can also be HETATM (if a residue is non-standard, like MSEe)

[wojdyr]

In gemmi and Servalcat we use a more dsophisticated heuristic:

The heuristic takes into account

residue types (peptide, nucleotide, or other), record types (ATOM/HETATM),
distances between consecutive residues, and gaps in numbering. This allows
us, in almost all cases, to determine the end of the polymer.

[glevans]

Is there an option in Gemmi to add TER records using this heuristic logic?

I have looked through the documentation but did not spot it.

[wojdyr]

there is an option ignore_ter to ignore TER records from the input file.
and function st.setup_entities() that determines where chains end,among other thing.
Then, if you write a pdb file, it will have TER records determined by gemmi.

[tdudgeon]

@wojdyr I don't see an ignore_ter option: https://project-gemmi.github.io/python-api/#read_pdb

[wojdyr]

python-api docs are outdated and can'tbe easily updated, as noted at https://project-gemmi.github.io/python-api/
you can either check the main doc:
https://gemmi.readthedocs.io/en/latest/search.html?q=ignore_ter&check_keywords=yes
of query the Python module:

$ PYTHONPATH=$HOME/gemmi/gemmi/build/py python3 -m pydoc gemmi.read_pdb |cat
Help on nb_func in gemmi:

gemmi.read_pdb = <nanobind.nb_func object>
    read_pdb(filename: str, max_line_length: int = 0, ignore_ter: bool = False, split_chain_on_ter: bool = False) -> gemmi.Structure

``