Recent Releases of randepict
randepict - v1.2.0
1.2.0 (2023-07-05)
Features
- add Release-please for automated releases (cb8a74f)
- orderly enumeration of R group indices (0481aaa)
Bug Fixes
- Failing documentation theme issues (987db2c)
Documentation
- version bump: 1.1.8-->1.2.0 (9c54d7c)
- Python
Published by github-actions[bot] over 2 years ago
randepict - 1.1.7
- Major restructuring of the code (this addresses https://github.com/OBrink/RanDepict/issues/5)
- 2D coordinate layouts generated with Indigo are now randomly rotated
- Users now can generate depictions along with cxSMILES strings that contain dimensionless coordinates that describe the positions of the atoms as they have been depicted
- XY-shearing has been removed from the augmentation lists to not to influence the coordinates
- Only the CDK, RDKit and Indigo are used here as PIKAChU re-generates coordinates during the depiction process
- Python
Published by OBrink over 2 years ago
randepict - RanDepict 1.1.5
- the CDK depictions are generated using more fonts for labels
- alkyl chain abbreviations are more diverse (eg. an ethyl group is now also abbreviated as "C2H5" instead of just as "Et")
- "Y" has been added as an R group variable per default for generating Markush SMILES using the RandomMarkushStructureCreator. We are aware of the conflict with the element Yttrium, but it appears that the R group variable is a typical case, while the occurrence of Yttrium is very rare A user who does not want to use it can still exclude it.
- The RandomMarkushStructureCreator now also includes the lowercase characters a-f in R group labels. For example, labels like "R1a" can be generated.
- The RandomMarkushStructureCreator does not replace isotopes or charges atoms with R group labels anymore as the resulting SMILES are invalid (see #35).
- The JVM heap size has been set to 4GB. We hope that this helps with #25, but this needs to be validated. For now, it does not break anything.
- Python
Published by OBrink about 3 years ago
randepict - RanDepict 1.1.4
- implemented a workaround for a bug in PIKAChU that leads to problems when depicting large molecules
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.1.3
- Fixed problem with Pillow's resampling method being accessed differently in different versions
- Avoid replacing chiral R groups in RandomMarkushStructureCreator as it leads to the generation of invalid SMILES
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.1.2
What's Changed
- Tm dev by @OBrink in https://github.com/OBrink/RanDepict/pull/26:
- Fixed broken tests
- Updated to CDK 2.8
- enable configuring depictor styles by @tulay in https://github.com/OBrink/RanDepict/pull/31
- RandomDepictor objects can now be instantiated using a configuration
- Configuration class added
- Tests added
New Contributors
- @tulay made their first contribution in https://github.com/OBrink/RanDepict/pull/31
Full Changelog: https://github.com/OBrink/RanDepict/compare/1.1.1...1.1.2
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.1.1
- Depiction feature fingerprints are now implemented for the PIKAChU depictions
- A bug with path definitions of the backgrounds for the hand-drawn-like augmentations has been fixed.
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.1.0
- Update script for tfrecord dataset generation
- Manual handling of processes for multiprocessing
- Dead processes are shut down automatically
- We previously had problems with multiprocessing's starmap function as processes froze and nothing happened after a while (when generating datasets of hundreds of millions of images). We have not really found the root of the problem here, as we do not know why the processes froze in the first place. It only started occurring with the generation of markush structure dataset with the implementation of PIKAChU. We need to take a closer look at this eventually. On the bright side - it works now and has been tested with the creation of a dataset of >500,000,000 markush structures.
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.0.9
- install PIKAChU from Pypi
- setup and tox updated for Python 3.10
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.0.8
What's Changed
- Add script for fingerprint-based generation of structure depiction datasets. The script generates a set of structure depictions and the same set of depictions with augmentations. We hope to be able to better analyse the effect of the augmentations on our OCSR results.
- Does not really belong here, but is relevant: PIKAChU did not close plots after returning an array that represents the structure depiction. This was fixed here.
- Add contribution statistics by @Kohulan in https://github.com/OBrink/RanDepict/pull/18
Full Changelog: https://github.com/OBrink/RanDepict/compare/1.0.7...1.0.8
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.0.7
Bugfix for generation of fingerprint-based bis datasets
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.0.6
- Added PIKAChU depiction functions
Added markush structure depictions
Depiction feature fingerprints for the above-mentioned features are still missing
- Python
Published by OBrink over 3 years ago
randepict - RanDepict 1.0.5
- Implementation of Feature Fingerprints to ensure diversity
- Addition of documentation, tutorial and examples
- A lot of smaller fixes
- Python
Published by OBrink about 4 years ago
randepict - RanDepict 1.0.3
- Fixed broken selection of random depiction options with Indigo
- Indigo depictions are no longer saved in temporary files
- Python
Published by OBrink over 4 years ago
randepict - RanDepict v1.0.2
Changelog - Fixed error when accidentally checking truth value of an array. - Handle problem with RDKit not being able to read some SMILES. - Minor bug fixes. - Deployed on PyPI.
- Python
Published by Kohulan over 4 years ago
randepict - RanDepict 1.0.0
This is the first release RanDepict, an easy-to-use utility to generate a big variety of chemical structure depictions (random depiction styles and image augmentations) based on RDKit, CDK and Indigo. We hope to contribute to facilitating the creation of diverse training datasets for deep learning-based OCSR tools.
- Python
Published by OBrink over 4 years ago