Release_2024.03.1

(Changes relative to Release_2023.09.1)

Acknowledgements

(Note: I'm no longer attempting to manually curate names. If you would like to see your contribution acknowledged with your name, please set your name in GitHub)

Mark Archibald, Armin Ariamajd, Chris Von Bargen, Jason Biggs, Jonathan Bisson, Jan C. Brammer, Jessica Braun, Benoît Claveau, David Cosgrove, James Davidson, Hussein Faara, Théophile Gaudin, Gareth Jones, Christoph Hillisch, Tad Hurst, Kevin Keating, Brian Kelley, Joos Kiener, David Lounsbrough, Jeremy Monat, Dan Nealschneider, Yoshinobu Ogura, Marta Pasquini, Yakov Pechersky, Patrick Penner, Rachael Pirie, Ricardo Rodriguez-Schmidt, Nate Russell, Ivan Tubert-Brohman, Matthew Seddon, Leonid Stolbov, Paolo Tosco, Riccardo Vianello, Franz Waibl, Rachel Walker, sitanshubhunia, skystreet8, dehaenw, dhibbit, vslashg, nbehrnd, MarioAndWario, levineds-meta

Highlights

• An initial version of support for atropisomers has been added; this will be expanded in future releases.
• Support for using multiple threads has been added in a few more places: many operations in rdMolStandardize, the fingerprint generators, and GetBestRMS()/GetAllConformerBestRMS()
• The initial release of version 2 of the RDKit C++ API; we will continue to expand this in future releases. The new API makes it easier to write correct and memory safe code. The current API is still supported and will remain so for the forseeable future, but we encourage C++ developers to start using v2 of the API in their code.

Backwards incompatible changes

• Two changes to improve the defaults for conformer generation: the functions EmbedMolecule() and EmbedMultipleConfis() now use ETKDGv3 by default (previously they were using ETKDGV1) and only consider heavy atoms when calculating RMSD for conformer pruning (previously Hs were alos considered).
• The way that the number of radical electrons is calculated for atoms coming from mol blocks has been changed. Systems like a [CH] marked as a DOUBLET will now have three radical electrons assigned. This is consistent with the value from SMILES.
• The validation classes in MolStandardize were refactored in order to offer a simpler and more consistent API. In the C++ implementation, the MolVSValidations base class was removed and consolidated into ValidationMethod. Consequently, the validate method replaced run in the subclasses related to MolVS (namely NoAtomValidation, FragmentValidation, NeutralValidation, and IsotopeValidation) and all subclasses of ValidationMethod are now required to implement a copy method. Moreover, MolStandardize::ValidationErrorInfo was redefined as an alias for std::string. The changes related to the MolVS validation methods were similarly implemented in the Python API.
• Metal atoms (really any atom which has a default valence of -1) now have their radical electron count set to zero if they form any bonds. Metal atoms/ions without bonds will continue to be assigned a radical count of either 1 or 0 if they do/do not have an odd number of valence electrons. It is not possible in a cheminformatics system to generally answer what the spin state of a metal atom should be, so we are taking a simple and easily explainable approach. If you know the spin state of your species, you can directly provide that information by calling SetNumRadicalElectrons().
• Chirality will now be perceived for three-coordinate atoms with a T-shaped coordination environment and the wedge in the stem of the T. If we are perceiving tetrahedral stereo, it's possible to interpret this unambiguously.
• Bug fixes in the v2 tautomer hash algorithm will change the output for some molecules. Look at PR #7200 for more details: https://github.com/rdkit/rdkit/pull/7200
• RMS pruning during conformer generation now symmetrizes conjugated terminal groups by default. This can be disabled with the parameter "symmetrizeConjugatedTerminalGroupsForPruning"

New Features and Enhancements:

• Support writing detailed SMARTS queries to CTABs using the SMARTSQ mechanism (github issue #5819 from greglandrum)
• add more error checking to substance groups (github issue #5923 from greglandrum)
• add maxRecursiveMatches to SubstructMatchParameters (github issue #6017 from greglandrum)
• Removed some code duplication between Depictor.cpp and common.h (github pull #6799 from greglandrum)
• Add support for writing chirality and stereo in MaeWriter (github pull #6810 from rachelnwalker)
• Implement MinimalLib get_mcs() version that returns JSON (github pull #6812 from ptosco)
• support generalized substructure search in the SubstructLibrary (github pull #6835 from greglandrum)
• Support copying of GeneralizeQueryMolecules (github issue #6851 from greglandrum)
• Enable chemist-friendly depiction of R-groups (github pull #6866 from ptosco)
• Allow building DetermineBonds without YAeHMOP support (github pull #6885 from greglandrum)
• Add multithreading to getBestRMS and new getAllConformerBestRMS (github pull #6896 from greglandrum)
• switch to catch2 v3 (github pull #6898 from greglandrum)
• minilib functions exposure: mmpa (github pull #6902 from StLeonidas)
• atropisomer handling added (github pull #6903 from tadhurst-cdd)
• Add multi-threaded versions of some MolStandardize operations (github pull #6909 from greglandrum)
• Add (multithreaded) functions to the fingerprint generators for calculating multiple fingeprints in one call (github pull #6910 from greglandrum)
• Add Python modules to generate stubs and automatically patch docstrings (github pull #6919 from ptosco)
• Update molecular templates headers and drop bond-length tests (github pull #6960 from github-actions[bot])
• Add in place and multithread support for more of the MolStandardize code (github pull #6970 from greglandrum)
• Enable in-tree builds and improve overloaded constructor docstrings (github pull #6980 from ptosco)
• Change the defaults for the conformer generation to be ETKDGv3 (github pull #6985 from greglandrum)
• Added fingerprints to GeneralizedSubstruct search and extended SWIG wrappers (github pull #6991 from jones-gareth)
• Allow sanitization to be disabled in PandasTools.LoadSDF (github issue #7019 from christophhillisch)
• Add Atom::hasValenceViolation (Take 2) (github pull #7030 from cdvonbargen)
• Please consider exposing maxBondMatchPairs param in rdRascalMCES.RascalOptions() (github issue #7054 from nate-russell)
• Copy stereo and substance groups during insertMol (github issue #7064 from cdvonbargen)
• [v2 API] FileParsers (github issue #7074 from greglandrum)
• [v2 API] Reaction Parsers (github issue #7075 from greglandrum)
• Rationalize attachment points (github issue #7078 from cdvonbargen)
• refactoring of MolStandardize validation module (github pull #7085 from rvianello)
• Add a 'force' option to MolStandardizer::Uncharger (github pull #7088 from rvianello)
• support sanitization of reaction product templates (github pull #7095 from greglandrum)
• Support atropisomers in the conformer generator (github pull #7098 from greglandrum)
• Compatibility with pathlib.Path (github pull #7100 from PatrickPenner)
• Add option to sanitize reaction components like molecules (github issue #7108 from MartaPasquini)
• [v2 API] MRV parsers (github pull #7110 from greglandrum)
• Add v2 API for the molecule CDXML parser (github pull #7113 from greglandrum)
• Make addStereoAnnotation public (github issue #7140 from cdvonbargen)
• optimize batch operations when editing molecules (github pull #7145 from bp-kelley)
• V2 API for the MolSuppliers (github pull #7168 from greglandrum)
• Improve output of debugMol (github pull #7172 from greglandrum)
• update cookbook, draw molecule with atom indices (github pull #7173 from nbehrnd)
• Colinear bonds in depiction cause stereo to be lost when converting to mol block (github issue #7177 from mps-hlx)
• Update MinimalLib Dockerfiles (github pull #7182 from ptosco)
• allow perception of stereo from T-shaped structures (github pull #7183 from greglandrum)
• switch the TFD code to use a fingerprint generator (github pull #7187 from greglandrum)
• Don't reset computed properties if already empty (github pull #7188 from rachelnwalker)
• Enhance molzip to properly handle RGroupDecompositions (github pull #7202 from bp-kelley)
• Add some ExplicitBitVect operations to Swig (github pull #7204 from jones-gareth)
• Some modernization of core GraphMol classes (github pull #7228 from greglandrum)
• Custom decimal precision (github pull #7229 from PatrickPenner)
• Add Double Cubic Lattice Volume (DCLV). (github pull #7234 from RPirie96)
• feat(minilib): expose the options parameter in get_inchi (github pull #7240 from BenoitClaveau)
• Postpone clearing computed properties until after all Hs removed (github pull #7241 from rachelnwalker)
• Speed up cleanMolStereo (github pull #7244 from ricrogz)
• add HetAtomProtomerv2 (github pull #7253 from greglandrum)
• Support zero order bonds in V3K CTABs (github pull #7269 from greglandrum)
• add option to symmetrize conjugated terminal groups when RMS pruning conformers (github pull #7270 from greglandrum)

Bug Fixes:

• STEREOANY bonds lead to non-stable SMILES/SMARTS strings (github issue #5499 from ricrogz)
• Chemical reactions with radicals cannot be pickled and unpickled. (github issue #5890 from sitanshubhunia)
• Postgresql: exact search showing false with radicals from CXSMILES (github issue #6276 from sitanshubhunia)
• CXSMILES: atom with labels should not also have dummyLabel property set (github issue #6309 from greglandrum)
• Query Features: Different input format leads to a different molecule (github issue #6349 from kienerj)
• non-physical radical counts being preserved (github issue #6370 from greglandrum)
• MolEnumerator: use repeat counts for SRUs when present (github issue #6429 from greglandrum)
• Unexpected non-matching ElementGraph hashes (github issue #6472 from jepdavidson)
• Fixes for canonicalization, and stereochemistry (github pull #6743 from tadhurst-cdd)
• MCS query incorrect when ringCompare=RingCompare.StrictRingFusion (github issue #6773 from d-b-w)
• Fixes bug in get_sss_json() (github pull #6806 from ptosco)
• SWIG builds failing on Windows (github pull #6808 from jones-gareth)
• Double bonds should not be depicted as crossed bonds in the presence of wavy bonds (github issue #6816 from ptosco)
• We should be able to run the tests without boost::iostreams (github issue #6818 from greglandrum)
• Fix stereo bond corruption on RGD. (github pull #6832 from jones-gareth)
• MurckoScaffold.MakeScaffoldGeneric() has issues with isotopes (github issue #6836 from dehaenw)
• Fix unclosed resource in BuildFuncGroupHierarchy (github pull #6846 from ricrogz)
• RGD: Fix doEnumeration true for cores that are not bundles (github pull #6857 from jones-gareth)
• Fix build error when serialization is off. (github pull #6867 from vslashg)
• Wavy bonds in mol blocks can't be stereo enumerated (github issue #6876 from bp-kelley)
• CDXML read of AND1 group (specifying racemic center) gets associated into an OR1 group (github issue #6887 from pechersky)
• Segfault in JSONToMols when "commonchem" is an int (github issue #6890 from i-tub)
• reapplyMolBlockWedging() should retain ENDDOWNRIGHT, ENDUPRIGHT dirs (github issue #6893 from ptosco)
• MMPA FragmentMol segfaults when new stereo perception is turned on (github issue #6900 from jasondbiggs)
• PositionVariationOp::getVariationCounts() does unnecessary copies of vectors (github issue #6906 from whosayn)
• Obtaining descriptors via Descriptors.descList results in duplication of SPS. (github issue #6928 from wsuzume)
• Some Clang-specific build instructions skip some clang compilers on mac (github issue #6941 from whosayn)
• With new stereo, removing H from an Imine double bond does not remove bond stereo (github issue #6944 from ricrogz)
• FindMolChiralCenters should honor RDK_USE_LEGACY_STEREO_PERCEPTION (github issue #6945 from ricrogz)
• generateDepictionMatching2DStructure does not optimally align when refPatt!=None, allowRGroups=False, alignOnly=True (github issue #6952 from ptosco)
• SpacialScore ignores undefined bond stereo (github issue #6957 from jasondbiggs)
• GetAtomPairFingerprint yields different rooted FP from generator (github issue #6958 from ptosco)
• DetermineBonds() for PH3 yields no bonding (github issue #6961 from dhibbit)
• Highlights of triple bonds come out wrong (github issue #6968 from DavidACosgrove)
• MaeMolSupplier cannot read dummy atoms from Maestro files (github issue #6973 from ricrogz)
• Chem.FindMolChiralCenters function should not be sensitive to atom-map numbers (github issue #6975 from skystreet8)
• Parsing a Mol leaks the "_needsDetectBondStereo" property (github issue #6981 from ricrogz)
• SubstructMatch maxRecursiveMatches is not being honored (github issue #6983 from ricrogz)
• HierarchicalClusterPicker::pick() randomly fails with Invariant Violation (github issue #7001 from ricrogz)
• rdkit.Dbase doesn't work correctly with Python 3.12 (github issue #7009 from rvianello)
• "Inconsistent state" when manually sanitizing and assigning stereo when using the new stereo algorithm (github issue #7023 from ricrogz)
• Spacing bug in compute2DCoordsForReaction (github issue #7028 from KevKeating)
• Update distance bounds calculation for conjugated double bonds in macrocycles (github pull #7032 from fwaibl)
• Middle line in triple bond drawn to incorrect point when a wedged bond is present (github issue #7036 from greglandrum)
• fragmentation of mol loses any sgroups (github pull #7056 from tadhurst-cdd)
• CSharp Wrapper ExtendedQueryMol Read Access Violation (github issue #7069 from jones-gareth)
• removing an atom should not remove all stereo groups involving that atom. (github issue #7071 from greglandrum)
• Sanitizing and assigning stereo twice can change bond stereo with new stereo (github issue #7076 from ricrogz)
• testConrec.cpp:130 fails on ARM64 (github issue #7083 from bjonnh-work)
• Wrong stereochemistry in embedded rings from stereospecific SMILES (github issue #7109 from brje01)
• Quaternary nitrogens with hydrogens are not a candidate for stereo (github issue #7115 from bp-kelley)
• Some metal centers get radical electrons (github issue #7122 from cdvonbargen)
• AddHs sets "no implicit Hs" on the atoms were Hs are added (github issue #7123 from ricrogz)
• ReplaceBond may cause valence issues in specific edge cases (github issue #7128 from ricrogz)
• Adding Wedge/Dash bond neighboring a stereo double bond causes a Precondition Violation (github issue #7131 from ricrogz)
• Stereo Annotation Appears Incorrect (github issue #7157 from lounsbrough)
• Unexpected exact mass values are returned for radium and radon (github issue #7162 from markarchibald)
• Adding missing headers in ReactionParser.h (github pull #7163 from tgaudin)
• fix: add PandasTools support for pandas 2.2 (github pull #7165 from AAriam)
• Fix leaking Bonds on unmatched ring closures (github pull #7178 from ricrogz)
• fix a problem with tautomeric systems being extended too far (github pull #7200 from greglandrum)
• Fixes #7181 (github pull #7206 from greglandrum)
• Fix Uncharger applying to already neutralized perhalic groups (github pull #7211 from rvianello)
• Fix Chem.Randomize.py (github pull #7232 from JanCBrammer)
• SGroup fields without values cause weird properties (github issue #7246 from ricrogz)
• Remove duplicate entry in fragment descriptors (github pull #7249 from levineds-meta)
• RDKit fails to parse "M RAD" lines were radical is 0 (github issue #7256 from ricrogz)
• Writing StereoGroups to Mol files should break lines at 80 characters (github issue #7259 from ricrogz)
• Update ring fusion cache when needed (github pull #7274 from ptosco)
• Ring stereo in SMILES inverted after sanitization in molecule with fragments (github issue #7295 from greglandrum)

Cleanup work:

• Switch over to using pytest to run the python tests (github pull #5916 from greglandrum)
• Redundant variablehasCoreDummies in R-group decomposition code (github issue #6779 from MarioAndWario)
• cmake cleanup (github pull #6814 from greglandrum)
• Remove boost::regex support (github issue #6817 from greglandrum)
• remove the deprecated python implementation of MolStandardize (github pull #6819 from greglandrum)
• Update CI, remove some warnings (github pull #6882 from greglandrum)
• Deprecate some of the ancient python-based ML code (github pull #6891 from greglandrum)
• Remove boost::regex support #6817 (github pull #6913 from whosayn)
• Fix minimal build, allow building without boost::serialization (github pull #6932 from greglandrum)
• Drop unrequired zlib include which may break the windows build (github pull #6966 from ricrogz)
• Compile time and runtime deprecation warnings (github pull #7004 from greglandrum)
• New tests for specical query atoms and atropisomers (github pull #7010 from tadhurst-cdd)
• fix GCC 13.2 warnings about redundant move in return statement (github pull #7029 from rvianello)
• fix check of python version when updating Filters.cpp (github pull #7035 from rvianello)
• fix several warnings originating from the swig wrappers (github pull #7063 from rvianello)
• lock the versions of a bunch of stuff used in the CI builds (github pull #7082 from greglandrum)
• remove deprecated packages from rdkit.ML (github pull #7107 from greglandrum)
• require SWIG 4.1+ at cmake config time (github pull #7139 from rvianello)
• RGD code cleanup (github pull #7186 from ptosco)
• remove the broken Dbase.DbReport module (github pull #7227 from greglandrum)
• remove a bunch of std::endls (github pull #7233 from greglandrum)
• Avoid rebuilding FreeSASA at every build for no good reason (github pull #7245 from ptosco)

Code removed in this release:

• The python implementations of MolStandardize has been removed. Please use the implementation in rdkit.Chem.MolStandardize.rdMolStandardize instead.
• The rdkit.six module, a leftover from the days when we supported both python 2 and python 3, has been removed
• The RDKit implementation of standard machine learning algorithms has been removed. The affected packages include: rdkit.ML.Composite, rdkit.ML.DecTree, rdkit.ML.KNN, rdkit.ML.ModelPackage, rdkit.ML.NaiveBayes, rdkit.ML.Neural rdkit.ML.{Analyze,Screen,Grow,Build}Composite, rdkit.ML.CompositeRun, rdkit.ML.EnrichPlot
• The Dbase.DbReport package was no longer working and has been removed.

Deprecated code (to be removed in a future release):

• The PDBMolSupplier class has been deprecated and will be removed in the next release
• The legacy Python code for drawing molecules has been deprecated and will be removed in the next release. This includes the following modules in rdkit.Chem.Draw: aggCanvas, cairoCanvas, canvasbase, MolDrawing, mplCanvas, qtCanvas, spingCanvas; the functions Draw.MolToImageFile(), Draw.MolToMPL(), and Draw.MolToQPixmap(); the "canvas" argument to the function Draw.MolToImage(); and calling Draw.MolToFile() with imageTypes other than PNG or SVG,