2023_03_1 (Q1 2023) Release

[greglandrum]

Release_2023.03.1

(Changes relative to Release_2022.09.1)

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)

Michael Banck, Christopher Von Bargen, Jason Biggs, Jonathan Bisson, Jacob
Bloom, Shang Chien, David Cosgrove, Iren Azra Azra Coskun, Andrew Dalke, Eloy
Félix, Peter Gedeck, Desmond Gilmour, Mosè Giordano, Emanuele Guidotti, Tad
Hurst, Gareth Jones, Calvin Josenhans, Maria Kadukova, Brian Kelley, Joos
Kiener, Chris Kuenneth, Martin Larralde, Krzysztof Maziarz, Jeremy Monat, Michel
Moreau, Rocco Moretti, Lucas Morin, Dan Nealschneider, Noel O'Boyle, Vladas
Oleinikovas, Rachael Pirie, Ricardo Rodriguez-Schmidt, Vincent F. Scalfani,
Gregor Simm, Marco Stenta, Georgi Stoychev, Paolo Tosco, Kazuya Ujihara,
Riccardo Vianello, Franz Waibl, Rachel Walker, Patrick Walters,
'dangthatsright', 'mihalyszabo88', 'Deltaus', 'radchenkods',
'josh-collaborationspharma', 'jkh', 'yamasakih'

Highlights

• The 2D coordinate generation can now optionally use templates when working with complex ring systems. We will continue to improve this functionality in future releases.
• There's now a single function which allows you to calculate all available 2D descriptors for a molecule: Descriptors.CalcMolDescriptors()
• Support for working with organometallic molecules has improved: drawings of these structures are now better and there's new code for switching back and forth between dative and multi-center views of the bonding in systems like ferrocene.
• The fingerprint generator code has been improved and expanded with the idea of allowing user to switch entirely to the new code for the supported fingerprint types: Morgan, RDKit, topological torsion, and atom pairs.

Backwards incompatible changes

• The ring-finding functions will now run even if the molecule already has ring information. Older versions of the RDKit would return whatever ring information was present, even if it had been generated using a different algorithm.
• The ring-finding functions now no longer consider dative bonds as possible ring bonds by default. All of the ring-finding functions have a new optional argument includeDativeBonds which can be used to change this behavior
• Generating 2D coordinates no longer has the side effect of running ring finding on molecules.
• The canonical SMILES and CXSMILES generated for molecules with enhanced stereochemistry (stereo groups) is different than in previous releases. The enhanced stereochemistry information and the stereo groups themselves are now canonical. This does not affect molecules which do not have enhanced stereo and will not have any effect if you generate non-isomeric SMILES. This change also affects the output of the MolHash and RegistrationHash code when applied to molecules with enhanced stereo.
• The doIsomericSmiles parameter in Java and C# ROMol.MolToSmiles() now defaults to true (previously it was false), thus aligning to the C++ and Python behavior.
• Double bonds which are marked as crossed (i.e. bond.GetBondDir() == Bond.BondDir.EITHERDOUBLE) now have their BondStereo set to Bond.BondStereo.STEREOANY and the BondDir information removed by default when molecules are parsed or AssignStereochemistry() is called with the cleanIt argument set to True.
• The conformers generated for molecules with three-coordinate chiral centers will be somewhat different due to the fix for Conformer Generation Fails for three-coordinate Ns with specified stereo #5883.
• Molecules which come from Mol or SDF files will now always have the "_MolFileChiralFlag" property set to the value of the chiral flag in the CTAB. In previous versions the property was not set if the chiral flag was 0.

Bug Fixes:

• GetSubstructMatches uniquify and maxMatches don't work well together
  (github issue GetSubstructMatches uniquify and maxMatches don't work well together #888 from adalke)
• DrawRDKBits raised RDKit error when it applied to the compounds that contains imidazole.
  (github issue DrawRDKBits raised RDKit error when it applied to the compounds that contains imidazole. #2164 from yamasakih)
• MolFromMol2File: O.co2 atom type correctness check ignores phosphate groups
  (github issue MolFromMol2File: O.co2 atom type correctness check ignores phosphate groups #3246 from chmnk)
• Enhanced Stereo is lost when using GetMolFrags(m, asMols=True)
  (github issue Segfault with coordgen v3.0.0 #4845 from kienerj)
• Segfault with coordgen v3.0.0
  (github issue Segfault with coordgen v3.0.0 #4845 from lucasmorin222)
• Dative bond and alkali and alkaline earth metals
  (github issue Dative bond and alkali and alkaline earth metals #5120 from marcostenta)
• RGD Stereochemistry in decomposed structure is not copied to the matching core
  (github issue RGD Stereochemistry in decomposed structure is not copied to the matching core #5613 from jones-gareth)
• fp.ToList() fails for empty molecule
  (github issue fp.ToList() fails for empty molecule #5677 from baoilleach)
• SMILES and SMARTS parse bonds in a different order
  (github issue SMILES and SMARTS parse bonds in a different order #5683 from ricrogz)
• postgresql makefile needs to be updated to use c++17
  (github issue postgresql makefile needs to be updated to use c++17 #5685 from mbanck)
• Exception raised when reading very large SMILES file
  (github issue Exception raised when reading very large SMILES file #5692 from DavidACosgrove)
• Update warning message about aromaticity detection
  (github pull Update warning message about aromaticity detection #5696 from d-b-w)
• stop building catch_main when tests are disabled
  (github pull stop building catch_main when tests are disabled #5697 from greglandrum)
• Make PandasTools.RGroupDecompositionToFrame re-entrant
  (github pull Make PandasTools.RGroupDecompositionToFrame re-entrant #5698 from greglandrum)
• PandasTools.RGroupDecompositionToFrame() should call ChangeMoleculeRendering()
  (github issue PandasTools.RGroupDecompositionToFrame() should call ChangeMoleculeRendering() #5702 from greglandrum)
• MolDraw2D should automatically set bond highlight color when atom colors are changed
  (github issue MolDraw2D should automatically set bond highlight color when atom colors are changed #5704 from greglandrum)
• Use correct _WIN32 macro for checking Windows target
  (github pull Use correct _WIN32 macro for checking Windows target #5710 from giordano)
• Environment not set properly in chirality tests for MinGW builds
  (github pull Environment not set properly in chirality tests for MinGW builds #5711 from ptosco)
• windows.h header should be lowercase
  (github pull windows.h header should be lowercase #5712 from ptosco)
• Fixes bond index parsing for w/c/t/ctu labels in CXSMILES/CXSMARTS
  (github pull Fixes bond index parsing for w/c/t/ctu labels in CXSMILES/CXSMARTS #5722 from ricrogz)
• Fix a deprecation warning in pythonTestDirRoot
  (github pull Fix a deprecation warning in pythonTestDirRoot #5723 from ricrogz)
• allowNontetrahedralChiralty should be honored when reading/writing SMILES
  (github pull allowNontetrahedralChiralty should be honored when reading/writing SMILES #5728 from greglandrum)
• CFFI/MinimalLib fixes
  (github pull CFFI/MinimalLib fixes #5729 from ptosco)
• Allow setting custom FREETYPE_LIBRARY/FREETYPE_INCLUDE_DIRS through CMake
  (github pull Allow setting custom FREETYPE_LIBRARY/FREETYPE_INCLUDE_DIRS through CMake #5730 from ptosco)
• Missing update path for postgreSQL from 3.8 to 4.2
  (github issue Missing update path for postgreSQL from 3.8 to 4.2 #5734 from Deltaus)
• Avoid passing a NULL pointer to CanSmiles()
  (github pull Avoid passing a NULL pointer to CanSmiles() #5750 from ptosco)
• CDXML reader incorrectly sets stereo on crossed double bonds
  (github issue CDXML reader incorrectly sets stereo on crossed double bonds #5752 from baoilleach)
• R atom label information lost in molfile if not backed by a M RGP entry
  (github issue R atom label information lost in molfile if not backed by a M RGP entry #5763 from eloyfelix)
• Missing monomer labels when depicting MON SGroups
  (github issue Missing monomer labels when depicting MON SGroups #5767 from eloyfelix)
• Wrongly oriented SGroup bracket
  (github issue Wrongly oriented SGroup bracket #5768 from eloyfelix)
• Adjust LocaleSwitcher on Windows when RDK_BUILD_THREADSAFE_SSS not set
  (github pull Adjust LocaleSwitcher on Windows when RDK_BUILD_THREADSAFE_SSS not set #5783 from roccomoretti)
• KekulizationException in tautomer canonicalization
  (github issue KekulizationException in tautomer canonicalization #5784 from d-b-w)
• ChemicalReactionToRxnBlock ignores separateAgents if forceV3000 is true
  (github issue ChemicalReactionToRxnBlock ignores separateAgents if forceV3000 is true #5785 from jacobbloom)
• extend the allowed valences of the alkali earths
  (github pull extend the allowed valences of the alkali earths #5786 from greglandrum)
• Minimallib build (rdkit-js) not working for release 2022.09.2
  (github issue Minimallib build (rdkit-js) not working for release 2022.09.2 #5792 from MichelML)
• Remove dependency on MSVC runtime DLL in MinGW builds
  (github pull Remove dependency on MSVC runtime DLL in MinGW builds #5800 from ptosco)
• Update macOS target platform to 10.13
  (github pull Update macOS target platform to 10.13 #5802 from ptosco)
• R# atom label information lost in molfile if not handled by the RGP spec
  (github issue R# atom label information lost in molfile if not handled by the RGP spec #5810 from eloyfelix)
• Stop using recommonmark in the documentation
  (github issue Stop using recommonmark in the documentation #5812 from greglandrum)
• Properties with new lines can create invalid SDFiles
  (github issue Properties with new lines can create invalid SDFiles #5827 from bp-kelley)
• Allow building PgSQL RPM and DEB packages
  (github pull Allow building PgSQL RPM and DEB packages #5836 from ptosco)
• Additional output is incorrect when FP count simulation is active
  (github issue Additional output is incorrect when FP count simulation is active #5838 from ptosco)
• Explicit valence check SiPa13fails for certain SMILES
  (github issue Explicit valence check fails for certain SMILES #5849 from josh-collaborationspharma)
• Set emsdk path for freetype in emscripten builds
  (github pull Set emsdk path for freetype in emscripten builds #5857 from ptosco)
• DrawMorganBit fails by default
  (github issue DrawMorganBit fails by default  #5863 from eguidotti)
• Fix R# atom label information lost in molfile if not handled by the RGP spec #5810 in V2000 mol files.
  (github pull Fix #5810 in V2000 mol files. #5864 from eloyfelix)
• Chemical drawings should be automatically enabled on Colab
  (github pull Chemical drawings should be automatically enabled on Colab #5868 from kuelumbus)
• use enhanced stereo when uniquifying in SimpleEnum
  (github pull use enhanced stereo when uniquifying in SimpleEnum #5874 from greglandrum)
• Conformer Generation Fails for three-coordinate Ns with specified stereo
  (github issue Conformer Generation Fails for three-coordinate Ns with specified stereo #5883 from gncs)
• Fix documentation example for KeyFromPropHolder
  (github pull Fix documentation example for KeyFromPropHolder #5886 from gedeck)
• Allow unrecognized atom types when strictParsing=False
  (github pull Allow unrecognized atom types when strictParsing=False #5891 from greglandrum)
• DetermineBonds assigning methyl carbon as tetrahedral center
  (github issue DetermineBonds assigning methyl carbon as tetrahedral center #5894 from jasondbiggs)
• numpy.float is no longer supported and causes exceptions
  (github issue numpy.float is no longer supported and causes exceptions  #5895 from PatWalters)
• moldraw2DTest1 failure when building on aarch64
  (github issue moldraw2DTest1 failure when building on aarch64 #5899 from vfscalfani)
• DetermineBondOrders running out of memory on medium-sized disconnected structure
  (github issue DetermineBondOrders running out of memory on medium-sized disconnected structure #5902 from jasondbiggs)
• clear MDL Rgroup labels from core atoms when we aren't using them
  (github pull clear MDL Rgroup labels from core atoms when we aren't using them #5904 from greglandrum)
• Conformer generator produces strange structure for substituted butadiene
  (github issue Conformer generator produces strange structure for substituted butadiene #5913 from gncs)
• MHFPEncoder::Distance doesn't compute a (Jaccard) distance
  (github issue MHFPEncoder::Distance doesn't compute a (Jaccard) distance #5919 from althonos)
• AvalonTools: Avoid that trailing garbage pollutes the fmemopen buffer
  (github pull AvalonTools: Avoid that trailing garbage pollutes the fmemopen buffer #5928 from ptosco)
• "not" queries in molfiles get inverted
  (github issue "not" queries in molfiles get inverted #5930 from d-b-w)
• CalcTPSA() doesn't use options when caching
  (github issue CalcTPSA() doesn't use options when caching #5941 from greglandrum)
• Bad drawing of end points for dative bonds
  (github issue Bad drawing of end points for dative bonds #5943 from DavidACosgrove)
• Extremes of drawn ellipses not being calculated correctly.
  (github issue Extremes of drawn ellipses not being calculated correctly. #5947 from DavidACosgrove)
• Arrow heads of dative bonds are different sizes
  (github issue Arrow heads of dative bonds are different sizes #5949 from DavidACosgrove)
• stop caching ring-finding results
  (github pull stop caching ring-finding results #5955 from greglandrum)
• Wrong bond endpoint when connecting to wedge bond in 2D image
  (github issue Wrong bond endpoint when connecting to wedge bond in 2D image #5963 from stgeo)
• Tiny change to get demo.html to load in legacy browsers
  (github pull Tiny change to get demo.html to load in legacy browsers #5964 from ptosco)
• detect bad double bond stereo in conformer generation
  (github pull detect bad double bond stereo in conformer generation  #5967 from greglandrum)
• drawing code should not generate kekulization errors
  (github issue drawing code should not generate kekulization errors #5974 from greglandrum)
• Adjust expected test results for newer freetype versions
  (github pull Adjust expected test results for newer freetype versions #5979 from greglandrum)
• CanonicalRankAtomsInFragment example in the documentation is not reproducible
  (github issue CanonicalRankAtomsInFragment example in the documentation is not reproducible #5986 from chmnk)
• Exception in RegistrationHash for molecules with bad bond directions
  (github pull Exception in RegistrationHash for molecules with bad bond directions #5987 from d-b-w)
• Updated the GetMolHash docstring for accuracy
  (github pull Updated the GetMolHash docstring for accuracy #5988 from irenazra)
• Fix a problem with pickling molecules with more than 255 rings
  (github pull Fix a problem with pickling molecules with more than 255 rings #5992 from greglandrum)
• Support Python 3.11
  (github pull Support Python 3.11 #5994 from greglandrum)
• Incorrect disconnection of CC(=O)O[Mg]OC(=O)C
  (github issue Incorrect disconnection of CC(=O)O[Mg]OC(=O)C #5997 from DavidACosgrove)
• PostgreSQL autovacuum stuck when molecules with query features are stored in mol columns
  (github issue PostgreSQL autovacuum stuck when molecules with query features are stored in mol columns #6002 from mihalyszabo88)
• Remove and from C++ headers
  (github pull Remove and from C++ headers #6003 from d-b-w)
• [PH3] incorrectly recognized as potential stereo center
  (github issue [PH3] incorrectly recognized as potential stereo center #6011 from greglandrum)
• Potential nontetrahedral stereo is recognized when nontetrahedral stereo is disabled.
  (github issue Potential nontetrahedral stereo is recognized when nontetrahedral stereo is disabled. #6012 from greglandrum)
• MolEnumerator is not propagating query information to molecules
  (github issue MolEnumerator is not propagating query information to molecules #6014 from greglandrum)
• Reactions do not propagate query information to products
  (github issue Reactions do not propagate query information to products #6015 from greglandrum)
• Error rendering to very small canvas
  (github issue Error rendering to very small canvas #6025 from DavidACosgrove)
• Bad double bond drawn for collinear atoms
  (github issue Bad double bond drawn for collinear atoms #6027 from DavidACosgrove)
• Fix some minor leaks
  (github pull Fix some minor leaks #6029 from ricrogz)
• Cannot draw molecule which includes an atom with a [!#X] query (for any X)
  (github issue Cannot draw molecule which includes an atom with a [!#X] query (for any X) #6033 from ShangChien)
• FragmentOnBonds may create unexpected radicals
  (github issue FragmentOnBonds may create unexpected radicals #6034 from ricrogz)
• Calling MurckoScaffold on molecule causes bug in pickling
  (github issue Calling MurckoScaffold on molecule causes bug in pickling #6036 from dangthatsright)
• Bump maeparser and coordgen versions
  (github pull Bump maeparser and coordgen versions #6039 from ricrogz)
• enhanced stereo is still included in CXSMILES if isomericSmiles=False
  (github issue enhanced stereo is still included in CXSMILES if isomericSmiles=False #6040 from greglandrum)
• Issues with ACS1996 drawing mode on a small canvas
  (github issue Issues with ACS1996 drawing mode on a small canvas #6041 from DavidACosgrove)
• Cyclobutyl group in a macrocycle triggers a stereo center
  (github issue Cyclobutyl group in a macrocycle triggers a stereo center #6049 from cdvonbargen)
• stereogroups not combined when parsing CXSMILES
  (github issue stereogroups not combined when parsing CXSMILES #6050 from greglandrum)
• Regression in depicting molecules with MDL query atoms
  (github issue Regression in depicting molecules with MDL query atoms #6054 from ptosco)
• Do not include dative bonds in ring finding by default
  (github issue Bad drawing of ferrocene #6058 from DavidACosgrove)
• Remove check for ring information from Atom::Match
  (github pull Remove check for ring information from Atom::Match #6063 from fwaibl)
• Correct docstring for minFontSize.
  (github pull Correct docstring for minFontSize. #6066 from DavidACosgrove)
• Minor code cleanup
  (github pull Minor code cleanup #6101 from ptosco)
• Dummy atoms should not be considered to be metals for M and MH queries
  (github issue Dummy atoms should not be considered to be metals for M and MH queries #6106 from greglandrum)
• Drawing in ACS mode crops small images
  (github issue Drawing in ACS mode crops small images #6111 from DavidACosgrove)
• Drawing in ACS1996 mode throws ValueError: Bad Conformer Id if molecule has no coords
  (github issue Drawing in ACS1996 mode throws ValueError: Bad Conformer Id if molecule has no coords.  #6112 from DavidACosgrove)
• Single atom or queries with hydrogens shouldn't trigger warning in mergeQueryHs
  (github issue Single atom or queries with hydrogens shouldn't trigger warning in mergeQueryHs #6119 from bp-kelley)
• DetermineBonds fails for single H atom
  (github issue DetermineBonds fails for single H atom #6121 from gncs)
• MinimalLib: avoid that assignStereochemistry() fails when removeHs=true
  (github pull MinimalLib: avoid that assignStereochemistry() fails when removeHs=true #6134 from ptosco)
• Round-tripping a reaction through pickle changes the outputs from RunReactants
  (github issue Round-tripping a reaction through pickle changes the outputs from RunReactants #6138 from kmaziarz)
• RGD and enhanced stereochemistry
  (github issue RGD and enhanced stereochemistry #6146 from jones-gareth)
• MaeMolSupplier requires bond block
  (github issue MaeMolSupplier requires bond block #6153 from cdvonbargen)
• Incorrect doule bond drawing with MolDraw2DSVG
  (github issue Incorrect doule bond drawing with MolDraw2DSVG #6160 from radchenkods)
• BondDir not cleared from bonds that aren't stereoactive
  (github pull BondDir not cleared from bonds that aren't stereoactive #6162 from greglandrum)
• Crossed bond not correctly drawn
  (github issue Crossed bond not correctly drawn #6170 from ptosco)
• ReactionFromRxnBlock fails on bond with reacting center status set
  (github issue ReactionFromRxnBlock fails on bond with reacting center status set #6195 from jones-gareth)
• Possible regression in the atom/bond highlighting code
  (github issue Possible regression in the atom/bond highlighting code #6200 from ptosco)
• Updated README to build the PostgreSQL cartridge + bug fix
  (github pull Updated README to build the PostgreSQL cartridge + bug fix #6214 from ptosco)
• Atoms may get flagged with non-tetrahedral stereo even when it is not allowed
  (github issue Atoms may get flagged with non-tetrahedral stereo even when it is not allowed #6217 from ricrogz)
• Fix TorsionFingerprints for 15 membered rings
  (github pull Fix TorsionFingerprints for 15 membered rings #6228 from kazuyaujihara)
• Fix build warnings
  (github pull Fix build warnings #6235 from ricrogz)
• Tri-coordinate atom with implicit + neighbor H atom is found potentially chiral
  (github issue Tri-coordinate atom with implicit + neighbor H atom is found potentially chiral #6239 from ricrogz)
• DativeBondsToHaptic doesn't set _MolFileBondEndPts correctly.
  (github issue DativeBondsToHaptic doesn't set _MolFileBondEndPts correctly. #6252 from DavidACosgrove)
• Round-tripping ferrocene through HapticBondsToDatives loses drawing highlights.
  (github issue Round-tripping ferrocene through HapticBondsToDatives loses drawing highlights. #6253 from DavidACosgrove)
• Using Chiral Tag instead of CIPCode to ensure preservation of chirality in addHs
  (github pull Using Chiral Tag instead of CIPCode to ensure preservation of chirality in addHs #6259 from HalflingHelper)
• Update assignSubstructureFilters.py
  (github pull Update assignSubstructureFilters.py #6270 from OleinikovasV)
• deal with deprecated DataFrame.append method
  (github pull deal with deprecated DataFrame.append method #6272 from greglandrum)
• compile-time error with GCC 12.2.1 on Fedora 36
  (github issue compile-time error with GCC 12.2.1 on Fedora 36 #6274 from rvianello)
• Fix UnitTestPandasTools for running without pandas installed.
  (github pull Fix UnitTestPandasTools for running without pandas installed. #6299 from roccomoretti)
• Aromatic dashes look bad
  (github pull Aromatic dashes look bad #6303 from greglandrum)

Cleanup work:

• Do deprecations for the 2023.03 release
  (github pull Do deprecations for the 2023.03 release #5675 from greglandrum)
• run clang_format
  (github pull run clang_format #5676 from greglandrum)
• Cleanup work on documentation Makefile
  (github pull Cleanup work on documentation Makefile #5804 from greglandrum)
• Refactor RGD moving function implementations from header to source files
  (github pull Refactor RGD moving function implementations from header to source files #5958 from jones-gareth)
• Disable POPCNT on M1
  (github pull Disable POPCNT on M1 #6081 from bjonnh-work)
• Remove spurious full stops from warnings.
  (github pull Remove spurious full stops from warnings. #6124 from DavidACosgrove)
• Reformat Python code for 2023.03 release
  (github pull Reformat Python code for 2023.03 release #6294 from ricrogz)
• Reformat C/C++ code ahead of 2023.03 release
  (github pull Reformat C/C++ code ahead of 2023.03 release #6295 from ricrogz)

New Features and Enhancements:

• mol V3000: multicenter dative bond
  (github issue mol V3000: multicenter dative bond #5121 from marcostenta)
• add molecular filter examples
  (github pull add molecular filter examples #5647 from RPirie96)
• Use templates in RDKit coordinate generation
  (github pull Use templates in RDKit coordinate generation #5643 from rachelnwalker)
• add MACCS fp to the MinimalLib
  (github pull add MACCS fp to the MinimalLib #5707 from eloyfelix)
• Enable additional parameters in prepareAndDrawMolecule() and expose them to CFFI/MinimalLib
  (github pull Enable additional parameters in prepareAndDrawMolecule() and expose them to CFFI/MinimalLib #5731 from ptosco)
• add includeRedundantEnvironments support to GetMorganGenerator
  (github pull add includeRedundantEnvironments support to GetMorganGenerator #5732 from greglandrum)
• FingerprintGenerator refactoring
  (github pull FingerprintGenerator refactoring #5748 from greglandrum)
• Expose RDLog to SWIG wrappers
  (github pull Expose RDLog to SWIG wrappers #5749 from ptosco)
• Add a timeout protection for CIP calculations
  (github pull Add a timeout protection for CIP calculations #5772 from tadhurst-cdd)
• Expose getMolFrags in CFFI and MinimalLib
  (github pull Expose getMolFrags in CFFI and MinimalLib #5774 from ptosco)
• Get the wrappers working with SWIG 4.0
  (github pull Get the wrappers working with SWIG 4.0 #5795 from greglandrum)
• Update AvalonTools to version 2.0.4a
  (github pull Update AvalonTools to version 2.0.4a #5796 from ptosco)
• Add early example of drawing a molecule to Getting Started with the RDKit in Python
  (github pull Add early example of drawing a molecule to Getting Started with the RDKit in Python #5803 from bertiewooster)
• Enable get_molblock(details_json) from MinimalLib
  (github pull Enable get_molblock(details_json) from MinimalLib #5806 from ptosco)
• Improvements to PandasTools.SaveXlsxFromFrame
  (github pull Improvements to PandasTools.SaveXlsxFromFrame #5835 from ptosco)
• swap boost::tuple to std::tuple
  (github pull swap boost::tuple to std::tuple #5851 from greglandrum)
• Make it easy to calculate all 2D descriptors
  (github pull Make it easy to calculate all 2D descriptors #5892 from greglandrum)
• Introduces AvgIpc descriptor
  (github pull Introduces AvgIpc descriptor #5896 from greglandrum)
• Add SMILES to each group abbreviation in Cookbook
  (github pull Add SMILES to each group abbreviation in Cookbook #5908 from bertiewooster)
• Support SubstanceGroups and StereoGroups in JSON
  (github pull Support SubstanceGroups and StereoGroups in JSON #5909 from greglandrum)
• Add info about mergeHs to README.
  (github pull Add info about mergeHs to README. #5910 from DavidACosgrove)
• Cookbook - update entry 1 and add entries 38 and 39
  (github pull Cookbook - update entry 1 and add entries 38 and 39 #5918 from vfscalfani)
• Allow the sources of conformer generation failures to be retrieved
  (github pull Allow the sources of conformer generation failures to be retrieved #5960 from greglandrum)
• Create getExperimentalTorsions() function
  (github pull Create getExperimentalTorsions() function #5969 from greglandrum)
• Molblock wedging improvements
  (github pull Molblock wedging improvements #5981 from ptosco)
• MinimalLib JS functions to add/remove Hs in place
  (github pull MinimalLib JS functions to add/remove Hs in place #5984 from ptosco)
• Adds Pat Walter's Chembl Filters extraction to the FilterCatalog
  (github pull Adds Pat Walter's Chembl Filters extraction to the FilterCatalog #5991 from bp-kelley)
• Add depiction coordinates to molzip
  (github pull Add depiction coordinates to molzip #5993 from jones-gareth)
• Enable using STL algorithms on ROMol atoms and bonds
  (github pull Enable using STL algorithms on ROMol atoms and bonds #5995 from ptosco)
• Enable building MinimalLib as a plain JS file for usage in legacy/headless browsers
  (github pull Enable building MinimalLib as a plain JS file for usage in legacy/headless browsers #5999 from ptosco)
• Allow WriteSDF to create v3000 SDF files
  (github pull Allow WriteSDF to create v3000 SDF files #6004 from jkhales)
• add maxRecursiveMatches to SubstructMatchParameters
  (github issue add maxRecursiveMatches to SubstructMatchParameters #6017 from greglandrum)
• Expose fingerprint generator options to python
  (github pull Expose fingerprint generator options to python #6024 from greglandrum)
• Allow SMARTS of zero order bonds to match zero order bonds
  (github pull Allow SMARTS of zero order bonds to match zero order bonds #6037 from d-b-w)
• Change IUPAC metal->non-metal single bonds to dative
  (github pull Change IUPAC metal->non-metal single bonds to dative #6038 from DavidACosgrove)
• Add canonicalization of stereo groups (enhanced stereo)
  (github pull Add canonicalization of stereo groups (enhanced stereo) #6051 from greglandrum)
• Improve MaeMolSupplier API
  (github pull Improve MaeMolSupplier API #6053 from ricrogz)
• Enable optional visualization of complex query atoms in a more compact form
  (github pull Enable optional visualization of complex query atoms in a more compact form #6056 from ptosco)
• Start a Maestro file (.mae) writer.
  (github pull Start a Maestro file (.mae) writer. #6069 from ricrogz)
• Expose some stereochemistry-related functions to SWIG wrappers
  (github pull Expose some stereochemistry-related functions to SWIG wrappers #6075 from ptosco)
• Add option to only include shortest paths for topological torsion fingerprints
  (github pull Add option to only include shortest paths for topological torsion fingerprints #6090 from greglandrum)
• Enable "smilesSymbol" substitution in SMARTS
  (github pull Enable "smilesSymbol" substitution in SMARTS #6096 from ricrogz)
• Add the option to wedge two bonds at chiral centers
  (github pull Add the option to wedge two bonds at chiral centers #6108 from greglandrum)
• Another minor code cleanup
  (github pull Another minor code cleanup #6109 from ptosco)
• A few SWIG tweaks
  (github pull A few SWIG tweaks #6110 from ptosco)
• Stereochemistry-related SWIG updates
  (github pull Stereochemistry-related SWIG updates #6127 from ptosco)
• SWIG pickling improvements (and other cleanup)
  (github pull SWIG pickling improvements (and other cleanup) #6133 from ptosco)
• Improved handling of organometallics
  (github pull Improved handling of organometallics #6139 from DavidACosgrove)
• expose two missing QueryAtom types to python
  (github pull expose two missing QueryAtom types to python #6158 from greglandrum)
• Support Pseudoatoms like Pol and Mod in the RegistrationHash
  (github pull Support Pseudoatoms like Pol and Mod in the RegistrationHash #6175 from irenazra)
• Better name for areBondsLinear.
  (github pull Better name for areBondsLinear. #6196 from DavidACosgrove)
• add features to allow drawing molecules in arbitrary positions on a large canvas
  (github pull add features to allow drawing molecules in arbitrary positions on a large canvas #6210 from greglandrum)
• Support chirality when determining if a molecule is a reaction reactant
  (github issue Support chirality when determining if a molecule is a reaction reactant #6211 from jones-gareth)
• rdMolHash.MolHash function should allow customization of the CXSmiles via Chem.CXSmilesFields
  (github issue rdMolHash.MolHash function should allow customization of the CXSmiles via Chem.CXSmilesFields #6224 from irenazra)
• Updated README for cartridge installation into conda PostgreSQL
  (github pull Updated README for cartridge installation into conda PostgreSQL #6236 from ptosco)
• Add a function to translate the MDL chiral flag into enhanced stereo groups
  (github issue Add a function to translate the MDL chiral flag into enhanced stereo groups #6241 from ricrogz)
• Add support for generic matching in the PgSQL cartridge
  (github pull Add support for generic matching in the PgSQL cartridge #6269 from bjonnh-work)
• allowOptionalAttachments should also include terminal query atoms matching hydrogen
  (github pull allowOptionalAttachments should also include terminal query atoms matching hydrogen #6280 from ptosco)
• Exposed queryColour in MolDrawOptions
  (github pull Exposed queryColour in MolDrawOptions #6282 from ptosco)
• Add a new tautomer mol hash
  (github pull Add a new tautomer mol hash #6289 from glandrum)
• has_coords() now reports whether coords are 2D or 3D if present
  (github pull has_coords() now reports whether coords are 2D or 3D if present #6297 from ptosco)
• Improve the installation/testing instructions.
  (github pull Improve the installation/testing instructions for linux/conda #6298 from roccomoretti)

Code removed in this release:

• The SmilesParserParams option useLegacyStereo has been removed. Please use
  SetUseLegacyStereoPerception() instead.
• The following JS methods:
  • generate_aligned_coords()
  • get_morgan_fp()
  • get_morgan_fp_as_uint8array()
  • get_pattern_fp()
  • get_pattern_fp_as_uint8array()
    which used to take several individual parameters have been removed.
    Please use the versions which take a single JSON string parameter.
• The PrintAsBase64PNGString function in PandasTools has been removed.
  Please use PrintAsImageString instead.

---

This discussion was created from the release 2023_03_1 (Q1 2023) Release.

[adalke]

I see Torsion fingerprint generation changed between 2022_09 and 2023_03. The release notes say "Fix TorsionFingerprints for 15 membered rings", but CHEMBL2448329 with only a 9-membered ring, changed its fingerprint.

Overall there are 136 differences in ChEMBL 30 (starting from a previously generated SMILES file, not directly from the SDF distribution). All of them contain a boron. Almost all of them have a B-P ([CHEMBL4524079] (https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL4524079/), [CHEMBL4524080] (https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL4524080/), and CHEMBL4103340 do not).

Is this expected? Here's a reproducible, starting with 2022.09.5 (these all use Python 3.11 and Boost 1.82.0).

>>> import rdkit
>>> rdkit.__version__
'2022.09.5'
>>> from rdkit import Chem
>>> m = Chem.MolFromSmiles("BP(=O)(OC[C@H]1O[C@@H](n2cnc3c(N)nc(Cl)nc32)[C@H](O)[C@@H]1O)OP(=O)(O)OP(=O)(O)O")
>>> from rdkit.Chem import rdMolDescriptors
>>> fp = rdMolDescriptors.GetHashedTopologicalTorsionFingerprintAsBitVect(m)
>>> list(fp.GetOnBits())
[144, 156, 157, 236, 248, 280, 396, 432, 433, 436, 440, 448, 452, 453, 468, 496, 497, 500, 501, 520, 521, 552, 684, 685, 696, 772, 773, 800, 872, 873, 876, 1228, 1316, 1324, 1344, 1345, 1348, 1440, 1528, 1529, 1530, 1580, 1581, 1624, 1632, 1644, 1660, 1704, 1752, 1768, 1824, 1836, 1856, 1868, 1869, 1870, 1944, 1980, 2040]

Using 2023.03.1

>>> import rdkit
>>> rdkit.__version__
'2023.03.1'
>>> from rdkit import Chem
>>> m = Chem.MolFromSmiles("BP(=O)(OC[C@H]1O[C@@H](n2cnc3c(N)nc(Cl)nc32)[C@H](O)[C@@H]1O)OP(=O)(O)OP(=O)(O)O")
>>> from rdkit.Chem import rdMolDescriptors
>>> fp = rdMolDescriptors.GetHashedTopologicalTorsionFingerprintAsBitVect(m)
>>> list(fp.GetOnBits())
[144, 156, 157, 196, 236, 248, 280, 396, 432, 433, 436, 440, 448, 452, 453, 468, 496, 497, 500, 501, 520, 521, 552, 684, 685, 696, 772, 773, 872, 873, 876, 1228, 1316, 1324, 1344, 1345, 1348, 1356, 1440, 1528, 1529, 1530, 1580, 1581, 1624, 1632, 1644, 1660, 1752, 1768, 1824, 1836, 1856, 1868, 1869, 1870, 1944, 1980, 2040]

The differences:

>>> fp_old = [144, 156, 157, 236, 248, 280, 396, 432, 433, 436, 440, 448, 452, 453, 468, 496, 497, 500, 501, 520, 521, 552, 684, 685, 696, 772, 773, 800, 872, 873, 876, 1228, 1316, 1324, 1344, 1345, 1348, 1440, 1528, 1529, 1530, 1580, 1581, 1624, 1632, 1644, 1660, 1704, 1752, 1768, 1824, 1836, 1856, 1868, 1869, 1870, 1944, 1980, 2040]
>>> fp_new = [144, 156, 157, 196, 236, 248, 280, 396, 432, 433, 436, 440, 448, 452, 453, 468, 496, 497, 500, 501, 520, 521, 552, 684, 685, 696, 772, 773, 872, 873, 876, 1228, 1316, 1324, 1344, 1345, 1348, 1356, 1440, 1528, 1529, 1530, 1580, 1581, 1624, 1632, 1644, 1660, 1752, 1768, 1824, 1836, 1856, 1868, 1869, 1870, 1944, 1980, 2040]
>>> set(fp_new) - set(fp_old)
{196, 1356}
>>> set(fp_old) - set(fp_new)
{800, 1704}

The 136 differing records are:

CHEMBL2448329, CHEMBL4299778, CHEMBL4300169, CHEMBL4299873, CHEMBL2448332, CHEMBL2448333, CHEMBL4300689, CHEMBL4300879, CHEMBL122888, CHEMBL4301413, CHEMBL4301911, CHEMBL4302231, CHEMBL4302317, CHEMBL4302850, CHEMBL2448334, CHEMBL2448335, CHEMBL2448336, CHEMBL2448337, CHEMBL2448338, CHEMBL2448339, CHEMBL2086772, CHEMBL223397, CHEMBL223954, CHEMBL224851, CHEMBL4454496, CHEMBL2448345, CHEMBL2029004, CHEMBL2029005, CHEMBL388190, CHEMBL3350429, CHEMBL3350430, CHEMBL3350428, CHEMBL4524079, CHEMBL4524080, CHEMBL484082, CHEMBL4538289, CHEMBL4551891, CHEMBL1802095, CHEMBL1802094, CHEMBL601711, CHEMBL610595, CHEMBL611086, CHEMBL2028999, CHEMBL2029001, CHEMBL2029003, CHEMBL2373179, CHEMBL2029007, CHEMBL2029008, CHEMBL2029009, CHEMBL2029010, CHEMBL2029006, CHEMBL2021421, CHEMBL2029000, CHEMBL2029002, CHEMBL2086768, CHEMBL2086769, CHEMBL2086766, CHEMBL2086773, CHEMBL2068733, CHEMBL2086763, CHEMBL2086764, CHEMBL2086765, CHEMBL2086762, CHEMBL2068734, CHEMBL2086761, CHEMBL2181938, CHEMBL2181939, CHEMBL2181940, CHEMBL2181941, CHEMBL2181943, CHEMBL2181934, CHEMBL2181942, CHEMBL2309024, CHEMBL2368436, CHEMBL2368437, CHEMBL2368438, CHEMBL2368440, CHEMBL2368441, CHEMBL2368442, CHEMBL2368443, CHEMBL2368444, CHEMBL2368445, CHEMBL2364572, CHEMBL2374533, CHEMBL2386492, CHEMBL2386496, CHEMBL2386497, CHEMBL2386498, CHEMBL2386499, CHEMBL2386342, CHEMBL2386493, CHEMBL2386494, CHEMBL2386495, CHEMBL2386341, CHEMBL2386489, CHEMBL2448386, CHEMBL2448387, CHEMBL2448390, CHEMBL2448391, CHEMBL2448397, CHEMBL2448398, CHEMBL2448399, CHEMBL2448400, CHEMBL2448442, CHEMBL2448443, CHEMBL2448444, CHEMBL2448445, CHEMBL2448446, CHEMBL3087162, CHEMBL3087163, CHEMBL3087164, CHEMBL3087165, CHEMBL3087158, CHEMBL3087159, CHEMBL3087160, CHEMBL3087161, CHEMBL3087166, CHEMBL3087167, CHEMBL3392138, CHEMBL4299192, CHEMBL4299194, CHEMBL4299198, CHEMBL4299199, CHEMBL4299204, CHEMBL4299205, CHEMBL4299206, CHEMBL4299207, CHEMBL4299208, CHEMBL4299209, CHEMBL3603848, CHEMBL3604561, CHEMBL3604558, CHEMBL3604560, CHEMBL3558885, CHEMBL3558886, CHEMBL4103340.

[greglandrum]

Hi @adalke
This change is due to PR #5748, which changed the "legacy" fingerprinting functions to use the fingerprint generators under the hood.
In the case of the topological torsion fingerprint I didn't actually expect that to change anything, so I will look into this.

[adalke]

Interesting. My working model was that chemistry perception changed since I couldn't figure out why else seemingly only boron was affected. It's hard for me to read the C++ code, but nothing stands out eyeballing a diff of old/new TopologicalTorsionGenerator.cpp.

[adalke]

Torsional fingerprints using includeChirality=1 have changed quite a bit between 2022_09_5 and 2023_03_1. Which of the above items lead to that change?

Here's a reproducible (it occurs with both SDF and SMILES input). Both are with Python 3.11 and Boost 1.82.0:

>>> import rdkit
>>> rdkit.__version__
'2022.09.5'
>>> from rdkit import Chem
>>> m = Chem.MolFromSmiles("Cc1cc(C)n(CC(=O)NNC(=O)C[C@H]2Oc3ccccc3NC2=O)n1")
>>> from rdkit.Chem import rdMolDescriptors
>>> fp = rdMolDescriptors.GetHashedTopologicalTorsionFingerprintAsBitVect(m, includeChirality=1)
>>> list(fp.GetOnBits())
[12, 28, 48, 49, 136, 336, 376, 377, 412, 540, 544, 612, 672, 688, 744, 764, 765, 812, 868, 1028, 1112, 1120, 1148, 1204, 1268, 1288, 1292, 1296, 1380, 1496, 1632, 1712, 1740, 1741, 1744, 1745, 1780, 1812, 1816, 1844, 1864, 1872, 1884, 1885, 1996, 1997, 2012, 2036]

>>> import rdkit
>>> rdkit.__version__
'2023.03.1'
>>> from rdkit import Chem
>>> m = Chem.MolFromSmiles("Cc1cc(C)n(CC(=O)NNC(=O)C[C@H]2Oc3ccccc3NC2=O)n1")
>>> from rdkit.Chem import rdMolDescriptors
>>> fp = rdMolDescriptors.GetHashedTopologicalTorsionFingerprintAsBitVect(m, includeChirality=1)
>>> list(fp.GetOnBits())
[12, 28, 48, 49, 332, 336, 376, 377, 412, 488, 540, 544, 588, 612, 688, 744, 764, 765, 860, 1028, 1056, 1060, 1112, 1113, 1148, 1236, 1264, 1268, 1288, 1292, 1380, 1381, 1428, 1448, 1496, 1712, 1740, 1741, 1744, 1745, 1780, 1812, 1872, 1884, 1956, 1996, 1997, 2036]

Here are the differences in the bits:

>>> old_fp = [12, 28, 48, 49, 136, 336, 376, 377, 412, 540, 544, 612, 672, 688, 744, 764, 765, 812, 868, 1028, 1112, 1120, 1148, 1204, 1268, 1288, 1292, 1296, 1380, 1496, 1632, 1712, 1740, 1741, 1744, 1745, 1780, 1812, 1816, 1844, 1864, 1872, 1884, 1885, 1996, 1997, 2012, 2036]
>>> new_fp = [12, 28, 48, 49, 332, 336, 376, 377, 412, 488, 540, 544, 588, 612, 688, 744, 764, 765, 860, 1028, 1056, 1060, 1112, 1113, 1148, 1236, 1264, 1268, 1288, 1292, 1380, 1381, 1428, 1448, 1496, 1712, 1740, 1741, 1744, 1745, 1780, 1812, 1872, 1884, 1956, 1996, 1997, 2036]
>>> set(new_fp) - set(old_fp)
{1056, 1956, 1060, 1381, 1448, 488, 332, 588, 1264, 1428, 1236, 1113, 860}
>>> set(old_fp) - set(new_fp)
{672, 1120, 1632, 868, 136, 1864, 812, 1296, 1204, 1844, 1816, 2012, 1885}

The includeChirality=0 fingerprints are identical between these two RDKit versions.

[greglandrum]

This is also related to #5748

[adalke]

Last one I noticed a change in the AtomPair fingerprint when using fromAtoms, but I didn't notice anything in the release notes about it.

Here's a test program which I'll use to show the issue. It starts by downloading and storing a copy of PubChem record 9425015 into 9425015.sdf on the local file system, then show that fromAtoms=[28] and [29] generate the same bits, but [28,29] in the 2023.03.1 release does not include a bit found in the 2022.09.5 release.

import pathlib
import rdkit
import urllib.request
from rdkit import Chem
from rdkit.Chem import rdMolDescriptors

# This will cache/load a file in the local filesystem!

filename = pathlib.Path("9425015.sdf")
def get_molblock():
    try:
        return filename.read_text()
    except Exception:
        pass
    with urllib.request.urlopen(
            "https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/CID/9425015/record/SDF/?record_type=2d&response_type=display") as f:
        molblock = f.read()

    print("Cached molblock to", filename)
    filename.write_bytes(molblock)
    return molblock

def get_mol():
    mol_block = get_molblock()
    return Chem.MolFromMolBlock(mol_block)

def compare(fromAtoms, got, expected):
    if got == expected:
        print(f"Bits for {fromAtoms} matches RDKit 2022.09.5")
        return
    got = set(got)
    expected = set(expected)
    print(f"Bits for {fromAtoms} differ from RDKit 2022.09.5!")
    print("   extra:", sorted(got - expected))
    print(" missing:", sorted(expected - got))
        
        
print("Version:", rdkit.__version__)

m = get_mol()

for fromAtoms, expected in (
        ([28], [40, 124, 244, 356, 468, 472, 500, 501, 504, 744, 756, 760, 761, 880, 1016, 1020, 1140, 1224, 1228, 1240, 1244, 1248, 1312, 1480, 1496, 1497, 1544, 1988, 1992]),
        ([29], [220, 240, 241, 376, 500, 501, 616, 636, 708, 756, 764, 880, 976, 988, 1004, 1008, 1016, 1060, 1244, 1245, 1312, 1480, 1496, 1497, 1728, 1732, 1828, 1988, 2020]),
        ([28, 29], [40, 124, 220, 240, 241, 244, 356, 376, 468, 472, 500, 501, 502, 504, 616, 636, 708, 744, 756, 757, 760, 761, 764, 880, 881, 976, 988, 1004, 1008, 1016, 1017, 1020, 1060, 1140, 1224, 1228, 1240, 1244, 1245, 1248, 1312, 1313, 1480, 1481, 1496, 1497, 1498, 1544, 1728, 1732, 1828, 1988, 1989, 1992, 2020]),
        ):
    
    fp = rdMolDescriptors.GetHashedAtomPairFingerprintAsBitVect(m, fromAtoms=fromAtoms)
    compare(fromAtoms, list(fp.GetOnBits()), expected)

With 2022.09.5:

Version: 2022.09.5
Bits for [28] matches RDKit 2022.09.5
Bits for [29] matches RDKit 2022.09.5
Bits for [28, 29] matches RDKit 2022.09.5

With 2023.03.1, the [28,29] case does not generate bit 1498.

Version: 2023.03.1
Bits for [28] matches RDKit 2022.09.5
Bits for [29] matches RDKit 2022.09.5
Bits for [28, 29] differ from RDKit 2022.09.5!
   extra: []
 missing: [1498]

[greglandrum]

Also #5748

[adalke]

Thanks. My reading of #5748 was that it was refactoring and updating only, and not meant to change existing fingerprints.

I've added "switch to new fingerprint API" in the chemfp tracker. https://todo.sr.ht/~dalke/chemfp/13

[greglandrum]

Thanks. My reading of #5748 was that it was refactoring and updating only, and not meant to change existing fingerprints.

Yeah; I didn't expect that either. Otherwise it would have been mentioned in the "backwards incompatible changes" section of the release notes.

I do plan to track down what's going on and will update this thread after I've done so