Meeko

What's Changed
* Update classifier by fabaff in https://github.com/forlilab/Meeko/pull/119

New Contributors
* fabaff made their first contribution in https://github.com/forlilab/Meeko/pull/119

**Full Changelog**: https://github.com/forlilab/Meeko/compare/v0.5.0...v0.5.1

**What's new**
- Support for reactive docking ([chemrxiv](https://chemrxiv.org/engage/chemrxiv/article-details/6466fa77f2112b41e9c960c5), [instructions](https://github.com/forlilab/Meeko#reactive-docking))
- Easier to define new atom types from SMARTS 56
- Minimal support for receptor preparation ([instructions](https://github.com/forlilab/Meeko#2-prepare-protein-pdbqt))
- [Modified API](https://github.com/forlilab/Meeko#api-changes-in-v05)

**Fixes and other improvements**
- refuses to write PDBQT with Nan and Inf charges by default 55
- wildcards accepted by `mk_export.py`, e.g. `mk_export.py *.dlg` 43
- option `-c` in `mk_export.py` writes SDF with top pose from each cluster from autodock-gpu 38
- set grid box size and center based on existing ligand efb14bc532170491fcbf66fcaf6f177cfab97a98
- handle H isotopes 39
- `mk_export.py` reads gzipped files
- fix bond typing in `PDBQTMolecule` for single atom molecules 57
- other minor fixes and improvements

What's new
- Proper parsing of flexible sidechains and dockings with multiple simultaneous ligands
- Documentation for covalent docking. It consists of docking ligands as flexible sidechains, so translation and rotation do not take place. The only degrees of freedom are the rotatable bonds, and the input ligand is aligned with the protein by Meeko. This feature was present in `v0.3.3`.
- Renamed script `mk_copy_coords.py` to `mk_export.py`
- Changed license to LGPL-2.1

Fixes and other improvements
- Molecules with implicit hydrogens are skipped, warn if conformer not 3D
- API changes to `RDKitMolCreate`
- Fixed `flexible_amides=True` and `PDBQTMolecule.__iter__`
- Write docking energies to SDF

Fixes bug that causes small cycles, such as 6-member rings, to be modeled flexibly without glue atoms (e.g. CG0/G0).

Example: O1CC2C(CC1)C1C(C2)CCCC1 [sdf](https://github.com/forlilab/Meeko/blob/3e21ed0457fd330213f71e899f894da91817a234/test/small_cycle_data/tricycle-6-5-6.sdf)

First reported in https://github.com/ccsb-scripps/AutoDock-Vina/issues/106

New feature: support for covalent docking

Another macrocycle bugfix.

Previously, rings larger than `max_ring_size` were added to the `rigid_rings` list. Bonds that belong to rings in the `rigid_rings` list are prevented from being broken. For large macrocycles, such as vancomycin, some rings larger than `max_ring_size` encompass smaller rings that must be broken. The bugfix consists in not adding rings larger than `max_ring_size` to the `rigid_rings list`. The `rigid_rings` list is exclusive for rings smaller than `min_ring_size`.

The behavior of `max_ring_size` in unchanged from the user perspective: rings larger than the specified value will not be opened. This is because rings larger than `max_ring_size` are not listed as breakable rings.

This bug came into existence after we started using an exhaustive set of rings instead of a smallest set of smallest rings (SSSR), after migrating from OpenBabel to RDKit.

Macrocycle related bug fixes that arose from differences in ring perception between OpenBabel and RDKit.

Since RDKit does not return an SSSR (smallest set of smallest rings), some macrocycles were not opened. For example, for "c1cc2ccc1CCCCCCO2", each aliphatic carbon would be perceived as belonging to two rings, which would not happen under SSSR because benzene is the smallest ring. Then, no bonds would be broken because the algorithm refused to break bonds belonging to more than a single SSSR ring. The ring perception is now performed by an external algorithm (Hanser, Jauffret and Kaufmann) and the procedure to define the bonds to break was redesigned to work with a exhaustive set of rings instead of an SSSR.


Also the --pH command line option was removed from `mk_prepare_ligand.py`.