PolyMetriX v0.2.0

PolyMetriX v0.2.0

New Features

Polymer Class Splitter

• Added dedicated splitter for polymer classes enabling precise data partitioning based on polymer types
• Joins existing splitting strategies (Random, LOCOCV, Tg-Based)

Terminal Group Functionality

• Support for incorporating terminal groups on side chains and backbones
• Quantifies interaction potentials including polarity, reactivity, and chemical contributions

Comparator Module

• Improved mixture analysis capabilities for polymer-organic systems
• Calculates aligned featurizers for polymer-organic mixtures
• Supports compatibility assessments and predictive modeling for polymer-drug and polymer-solvent formulations

Improvements

• Featurizers: Refined with explicit terminal group integration
• Polymer-organic mixtures: Strengthened integration of molecular components (SMILES) with polymers
• Benchmarking: Maintains performance advantages against PolyBERT and Morgan fingerprints
• Dataset: 7,367 curated Tg data points optimized for new features
• Compatibility: Full support for Scikit-learn and PyTorch workflows

Technical Details

The release focuses on practical enhancements for polymer informatics workflows, particularly improving model robustness and interaction predictions in digital polymer chemistry applications.

See the updated API reference and documentation for implementation details.

v0.1.0

PolyMetriX v0.1.0 – First Release 🎉

We are excited to announce the first release of PolyMetriX, a comprehensive Python library designed to support digital polymer chemistry and polymer informatics workflows. 🚀

🔬 What is PolyMetriX?

PolyMetriX is an open-source machine learning framework for polymer informatics, offering:

• Feature engineering tools for polymers, including chemical and topological featurizers.
• A curated dataset of 7,367 polymer glass transition temperature (Tg) data points.
• A unified pipeline for data preparation, featurization, and model training.
• Custom splitting strategies for improved model generalization in polymer ML.

🌟 Key Features

✅ Advanced Featurizers:

• Extracts polymer representations at hierarchical levels (polymer, backbone, and side-chain).
• Supports Morgan fingerprints, RDKit molecular descriptors, and polymer-specific features.

✅ Curated Tg Dataset:

• 7,367 high-quality Tg data points .
• Provides a standardized benchmark for polymer property prediction.

✅ Robust Data Splitters:

• Random Splitting (for standard ML tasks).
• Leave-One-Cluster-Out Cross-Validation (LOCOCV) (for unbiased generalization tests).
• Tg-Based Splitting (for property-driven extrapolation models).

✅ Machine Learning Workflow:

• Supports ML-ready polymer featurization & dataset handling.
• Compatible with Scikit-learn, PyTorch, and other ML libraries.