SMiPoly

SMiPoly generates virtual libraries of potentially synthesizable polymers from small organic molecules using rule-based polymerization reactions to support de novo polymer design.

Key Features:

• Implementation and Modules: Python-based implementation with two submodules, monc.py and polg.py, that transform small organic molecules into polymer structures.
• Rule-Based Generation: Employs 22 chemical rules corresponding to commonly used polymerization reactions to produce chemically plausible polymers.
• Supported Polymer Types: Generates polymers across seven types: polyolefin, polyester, polyether, polyamide, polyimide, polyurethane, and polyoxazolidone.
• High-Throughput Generation: Produced 169,347 unique polymers from an initial set of 1,083 readily available monomers in the referenced dataset.
• Coverage and Novelty Analysis: Output compared with ~16,000 real-world synthesized polymers showed 48% coverage and 53% novelty.
• Exhaustive Library Output: Outputs a comprehensive library of potentially synthesizable polymers for downstream analysis.

Scientific Applications:

• De Novo Polymer Design: Generates candidate polymer structures for inverse molecular design workflows.
• Machine Learning Integration: Can be integrated with machine learning-assisted methodologies to aid identification of synthetic routes for designed polymers.
• Virtual Screening: Enables high-throughput virtual screening to streamline selection of synthesizable candidate polymers and accelerate polymer synthesis.

Methodology:

Form a monomer set from small organic molecules; use monc.py and polg.py to transform monomers and perform sequential polymerization reactions using 22 predefined chemical rules to generate an exhaustive list of polymers.