CReM

CReM implements a fragment-based structure generation framework that produces chemically valid structures with flexible control over diversity, novelty, synthetic complexity, and chemotypes of generated compounds.

Key Features:

• Chemically Valid Structure Generation: The framework generates structures that are chemically valid by design, avoiding invalid structures that may arise from deep learning models and conventional atom-based approaches.
• Synthetic Complexity Control: The framework explicitly addresses synthetic complexity of generated structures, unlike previous fragment-based approaches.
• Flexible Parameter Control: Users can adjust diversity, novelty, synthetic complexity, and chemotypes of generated compounds.
• Fragment-Based Approach: The framework uses fragment-based mutation to provide better novelty and diversity compared to conventional reaction-based approaches.
• Open-Source Implementation: The framework is implemented as a Python module for creating custom workflows for chemical space exploration.

Scientific Applications:

• De Novo Molecular Design: CReM supports de novo design studies where structure generator performance substantially influences outcomes.
• Chemical Space Exploration: Custom workflows can be created to explore chemical space for drug discovery and compound optimization.
• Synthetic Feasibility Optimization: The framework balances novelty and diversity with synthetic feasibility, addressing limitations of both reaction-based and atom-based approaches.

Methodology:

CReM uses a fragment-based mutation framework for structure generation.