MolPipeline: A Python Package for Processing Molecules with RDKit in Scikit-learn.

Area of Science:

• Cheminformatics
• Machine Learning
• Computational Chemistry

Background:

• Scikit-learn's Pipeline class facilitates machine learning workflows.
• Cheminformatics tasks often require custom data processing.
• Integrating these tasks into machine learning pipelines can be complex.

Purpose of the Study:

• Introduce MolPipeline, a Python package extending scikit-learn's Pipeline for cheminformatics.
• Enable automated, end-to-end cheminformatics pipelines scalable to large datasets.
• Improve handling of erroneous data instances within pipelines.

Main Methods:

• Wrap standard RDKit functionalities (e.g., SMILES I/O, descriptor calculation) within a pipeline framework.
• Develop building blocks for seamless integration of cheminformatics tasks.
• Incorporate features like scaffold splitting and molecular standardization.

Main Results:

• MolPipeline provides a user-friendly interface for building complex cheminformatics pipelines.
• The package effectively integrates RDKit functions into scikit-learn's ecosystem.
• Enhanced error handling for erroneous molecular instances is a key feature.

Conclusions:

• MolPipeline simplifies the creation of automated cheminformatics pipelines.
• It enhances the adaptability of machine learning workflows for diverse cheminformatics projects.
• The package facilitates efficient processing of large chemical datasets.