python-modelcif 1.3 release

- The new :class:`modelcif.CustomTemplate` class allows for custom templates
  (that have not been deposited in a database such as PDB) to be referenced,
  together with their atomic coordinates (#1).
- Model quality scores can now be defined that act on single features or
  pairs of features using the :class:`modelcif.qa_metric.Feature` and
  :class:`modelcif.qa_metric.FeaturePairwise` classes, respectively.
  Features can be defined as groups of atoms, residues, or asyms (#38).
- The :class:`modelcif.associated.QAMetricsFile` class should now be used
  to reference files that contain model quality scores. The old name
  (LocalPairwiseQAScoresFile) is deprecated. This allows for all types of
  QA scores, not just local pairwise scores, to be stored in a separate file.
- Sanity checks when writing out a file can now be disabled if desired,
  using the new ``check`` argument to :func:`modelcif.dumper.write`.
- :class:`modelcif.reference.TargetReference` now takes an ``is_primary``
  argument which can be used to denote the most pertinent sequence
  database reference.
- Information on model groups (:class:`modelcif.model.ModelGroup`) is now
  written to the new ``ma_model_group`` and ``ma_model_group_link`` mmCIF
  tables, instead of ``ma_model_list``, to match the latest ModelCIF
  dictionary. Old-style information in ``ma_model_list`` will still be
  used when reading a file if these new tables are missing.

python-modelcif 1.2 release

- Data that have been split over multiple mmCIF or BinaryCIF files can now
  be combined into a single :class:`ihm.System` object using the new
  ``add_to_system`` argument to :func:`modelcif.reader.read` (#10).
- A new example, ``associated.py``, has been added to demonstrate reading
  in data that has been split into multiple "associated" mmCIF files using
  :class:`modelcif.associated.CIFFile`.

python-modelcif 1.1 release

- The new class :class:`modelcif.model.NotModeledResidueRange` allows for
  the annotation of residue ranges that were explicitly not modeled.
  Any residue marked as not-modeled in all models will be excluded from
  the ``pdbx_poly_seq_scheme`` table.
- The ``util/make-mmcif.py`` script is now included in the installed package,
  so can be run if desired with ``python3 -m modelcif.util.make_mmcif``.
- The ``make_mmcif`` utility script will now automatically add any
  missing :class:`modelcif.model.NotModeledResidueRange` objects for
  not-modeled residue ranges.

python-modelcif 1.0 release

- Reference information in the ``struct_ref`` mmCIF table is now supported
  in addition to the ModelCIF-specific tables such as
  ``ma_target_ref_db_details``. :class:`modelcif.reference.TargetReference`
  now inherits from ``ihm.reference.Sequence`` and allows for the full
  database sequence, plus any differences between it and the modeled sequence,
  to be recorded. The ``align_begin`` and ``align_end`` arguments are now
  deprecated (#34).

python-modelcif 0.9 release

- Bugfix: :class:`modelcif.SoftwareGroup` now allows for parameters to
  be associated with each piece of software in the group, rather than
  with the group as a whole (#33).

python-modelcif 0.8 release

- :class:`modelcif.associated.File` now takes an optional ``data``
  argument to allow describing any modeling input/output that is stored
  in that file.
- RPM packages are now provided for Fedora and RedHat Enterprise Linux.

python-modelcif 0.7 release

- More examples have been added to demonstrate interconversion between
  mmCIF and BinaryCIF, and to validate mmCIF files.
- A utility script ``util/make-mmcif.py`` has been added which can add
  minimal ModelCIF-related tables to an mmCIF file, to add in deposition.
- The reader is now more robust when handling files that are not ModelCIF
  compliant (#31).
- The ``exptl`` table is no longer written to output mmCIF files, to conform
  with wwPDB's recommendation. Instead, the
  ``struct.pdbx_structure_determination_methodology`` data item denotes
  that the model is computational (#29).

python-modelcif 0.6 release

- :class:`ihm.ChemComp` now allows for custom chemical components to be
  defined in a chemical component dictionary (CCD) outside of the wwPDB CCD,
  such as the ModelArchive CCD, or in the file itself using descriptors such
  as SMILES or InChI in the :mod:`modelcif.descriptor` module.
- The ``ma_struct_assembly`` category is no longer written out to mmCIF
  files, as this is deprecated by ModelCIF (all models are required to
  have the same composition).
- Templates can now be described in AlphaFoldDB or PubChem using new
  :class:`modelcif.reference.TemplateReference` subclasses.
- HHblits e-values can now be used as alignment scores, using
  :class:`modelcif.alignment.HHblitsEValue`.
- Bugfix: :class:`modelcif.associated.CIFFile` now writes local files
  (if requested via ``categories`` or ``copy_categories``) even if it
  is placed inside a :class:`modelcif.associated.ZipFile` (#26).

python-modelcif 0.5 release

- A new class :class:`modelcif.ReferenceDatabase` allows describing
  collections of sequences that were used as part of the modeling protocol.
- Lists of ints or floats can now be given as software parameters to the
  :class:`modelcif.SoftwareParameter` class.

python-modelcif 0.4 release

- Sequence information for templates is now only written to template-specific
  categories in the mmCIF/BinaryCIF, not to the entity, entity_poly etc.
  tables, to properly comply with the ModelCIF dictionary.
- :class:`modelcif.Template` now takes a ``entity_id`` argument which can be
  used to provide the entity ID (if known) of the template in its own mmCIF
  file.
- External files (e.g. alignments, or quality scores) can now be referenced
  from the main file (using the :mod:`modelcif.associated` module). Selected
  CIF categories can automatically be written to these external files instead
  of the main file, in either mmCIF or BinaryCIF format (see
  :class:`modelcif.associated.CIFFile`).
- Non-polymer models can now be linked to their template using the
  :class:`modelcif.NonPolymerFromTemplate` class.
- Add classes for the PLDDT, PTM, and IpTM quality metrics.
- :class:`modelcif.reference.TargetReference` now supports the version
  and CRC64 checksum of the reference sequence.