03 Jan 12:57

TieuLongPhan

8abe152

Prepare release v.0.0.14

Changelog v0.0.14

Highlights

New Features & Enhancements

NXToGML Enhancement

Enhanced the NXToGML function to include an expanded representation of explicit hydrogens, along with a new strip function for hydrogen removal in context graphs.
- These improvements facilitate more efficient graph isomorphism operations by reducing computational expenses.

Chemical Conversion Updates

Updated conversion functions: gml_to_smat and smat_to_gml.
- Improvements include the ability to perform sanity checks while maintaining the atom map for hydrogens, enhancing the accuracy and utility of these conversions.

Rule Clustering Enhancement

Enhanced rule clustering capabilities to support nx.Graph formats and GML strings.
- This update allows for more versatile and robust rule application in graph transformations.

New PartialAAMs Method

Introduced a new method for expanding PartialAAMs using a gluing graph approach, which involves integrating the reaction center graph with reactants or product graphs.
- This innovative approach aids in constructing more accurate ITS structures, especially in complex chemical reactions.

Bug Fixes

GraphDescriptors Enhancement

Improved the GraphDescriptors function to address previously identified inaccuracies, ensuring more reliable and precise descriptor calculations.

Known Limitations

Limitations in PartialAAMs Methodology

Current Limitation: The new PartialAAMs method only functions correctly for balanced chemical reactions.
- Efforts are underway to extend this methodology to unbalanced reactions, aiming to broaden its applicability and effectiveness in diverse chemical scenarios.

Assets 2

05 Dec 08:59

TieuLongPhan

v0.0.13

a8a6d67

Prepare release v.0.0.13 Latest

Latest

Changelog v0.0.13

Highlights

New Features & Enhancements

GraphSignature Enhancement

Enhanced the GraphSignature function for improved graph matching and better representation of chemical reactions.
- The update includes optimizations that increase the efficiency of graph signature generation, ensuring more accurate results for complex ITS structures.

IUPAC Name Query Function (PubChem)

Added a function to query the IUPAC name of compounds directly from PubChem.
- This feature allows users to retrieve IUPAC names based on chemical structures.
- Note: This approach may experience time delays due to the querying process from PubChem.

Visualization Enhancement

Improved visualization capabilities by introducing the option to visualize rules in a manner similar to MØD.
- This enhancement provides users with a clearer representation of rule transformations similar to DPO diagram.
- Additionally, a variable for input title has been added to subplots, enabling custom titles for each visualization for improved clarity and organization.

Known Limitations

Time Delays in PubChem Querying

Current Bug: The query function for retrieving the IUPAC name from PubChem may experience time delays due to the querying process.
- This can affect the overall performance, especially when querying multiple compounds in a batch.

GraphSignature Performance

No proof that the GraphSignature function can be used as an alternative way for graph isomorphism checks.
- Users are advised to continue using established graph isomorphism methods for critical applications until further validation.

Assets 2

27 Nov 14:13

TieuLongPhan

v0.0.12

965ab58

Prepare release v.0.0.12

Changelog v0.0.12

Highlights

New Features & Enhancements

AAM Inference

Introduced AAM inference capabilities, allowing the generation of reaction SMILES with Atom-Atom Mapping (AAM) using rule-based transformations.
AAM generation is now integrated with rule transformation to predict and apply atom mappings for reaction steps.

PartialAAM Expansion

Enhanced PartialAAM expansion by refining the process of expanding partial Atom-Atom Mapped (AAM) reaction SMILES into full AAM RSMI (Reaction SMILES).
Improved expansion algorithm efficiency and accuracy, with additional normalization to handle redundant hydrogen mappings, ensuring that mappings are retained only for hydrogens in the reaction center.

AAM Validator

Added AAMValidator to verify the equivalence of AAMs between reaction steps.
This tool compares atom mappings across multiple reaction steps, ensuring that the mappings remain consistent and accurate throughout the transformation process.

SynVis Enhancement

Bug Fix: Fixed visualization issues in SynVis, ensuring correct coloring of bonds in ITS graphs.
- Green now represents new bonds being formed.
- Red indicates bonds that have been broken during the reaction.

Chemical Conversion Framework

Introduced the chemical_conversion function to support format conversion between different chemical representations.
- Now capable of converting between formats such as SMARTS, GML, and others, allowing users to seamlessly transform chemical data between multiple formats.

Multi-Step Reaction Framework (Prototype)

Developed a prototype for a multi-step reaction framework that allows users to input an original reaction along with a list of reaction rules and their application order.
- The framework generates the entire reaction sequence, step by step, with atom-atom mapping applied at each stage.

Known Limitations

Time Processing for DG Transformation

Current Bug: Time processing issues observed when transforming DG (Direct Graph) to SMILES with AAM.
- The transformation process can experience delays, particularly when handling complex reactions with extensive atom mappings.

MØD and RDKit Compatibility Issues (Ongoing)

Problem: Compatibility issues persist between MØD (conda version) and RDKit, resulting in Boost.Python.ArgumentError during RDKit operations.
Workaround:
- Add a workaround (using torch 2.0.1) before initializing MØD to mitigate compatibility issues.
- For pytest, use a fixed test order to avoid conflicts during testing workflows.

Assets 2

18 Nov 14:17

TieuLongPhan

v0.0.11

cfb522a

Prepare release v.0.0.11

Changelog v0.0.11

Highlights

New Features & Enhancements

MolToGraph and GraphToMol

Refactored source code for better modularity and maintainability.
Introduced a lightweight version of MolToGraph and GraphToMol for efficiency.
Added functionality to convert reaction SMILES into a graph representation.

Visualization

Enabled 2D coordinate generation from RDKit molecules for improved user-friendly MolGraph and ITSGraph visualizations.
Introduced a visualizer to process reaction SMILES with atom-atom mapping (AAM):
- Outputs separate visualizations for reactants, ITS graphs, and products.
Added the capability to save visualization outputs as PDFs for convenient sharing and documentation.

MØD Rule Application

Introduced the Transformation subpackage to support MØD-based graph transformations.
Enhanced compatibility with GitHub workflows by adding test cases for transformations.

PartialAAM Expansion (Preliminary Support)

Added the SynAAM subpackage for expanding PartialAAM reaction SMILES (RSMI) into full AAM RSMI.
Included a normalizer to reduce redundant hydrogen mappings and retain mappings only for hydrogens in the reaction center.
Implemented ITS decomposer to split ITS graphs back into reactant and product graphs.

Known Limitations

MØD and RDKit Compatibility Issues

Problem: Conflicts between MØD (conda version) and RDKit lead to Boost.Python.ArgumentError when using RDKit.
Workaround:
- Add a workaround (using torch 2.0.1) before initializing MØD to mitigate compatibility issues.
- For pytest, use a fixed test order to avoid conflicts during testing workflows.

Assets 2

11 Nov 09:02

TieuLongPhan

v0.0.10

8c17232

v0.0.10: Prepare release v.0.0.10 (#11)

Changelog v0.0.10

New Release Highlights

Improvements

Unified Graph Descriptors and Fingerprints
Combined the functionalities of graph descriptors and graph fingerprints into a cohesive module, simplifying the codebase and enhancing efficiency in graph property calculations and structural comparisons.
Parallel Processing for Descriptor Computation
Introduced parallel processing using joblib to handle graph descriptor computation across multiple entries simultaneously. This significantly reduces runtime on multi-core systems, particularly beneficial for large datasets. Users can specify the number of parallel jobs (n_jobs) and verbosity levels for flexible performance tuning.
MØD Rule Application (Preliminary)
Added preliminary support for applying MØD (Modeling Organic Databases) rules within the graph transformation pipeline. While not fully tested due to dependency conflicts in conda, this feature provides a framework for chemical rule applications in future updates.

Bug Fixes

Adapted MolToGraph for Unsanitized RDKit Molecules
Updated MolToGraph to handle unsanitized RDKit Mol objects, allowing more flexibility with RDKit molecule inputs without requiring pre-sanitization. This fix addresses compatibility issues and expands support for RDKit functionalities within the graph processing pipeline.

Assets 2