Research data for machine learning-guided structure–formulation–property modeling of DOPO-derived flame-retardant epoxy systems
Description
This dataset supports the manuscript “Machine Learning-Guided Structure–Formulation–Property Modeling and Virtual Screening of DOPO-Derived Flame-Retardant Epoxy Systems”. It contains processed literature-derived modeling data for DOPO-derived flame-retardant epoxy systems, final grouped cross-validation outputs, independent literature-based external-validation data, virtual-screening and candidate-prioritization results, and machine-readable source data supporting the principal tables and figures. The formal modeling database comprises 599 standardized experimental records curated from 130 published studies. Candidate-screening outputs include the formal 409-molecule screening library, applicability-domain-constrained predictions under 35 and 50 kW m−2 heat-flux scenarios, Pareto-screening results, and the final 19 cross-flux priority candidates. The associated version-controlled source code and reproducibility workflow are available from the DOPO-EP-ML GitHub repository, release v1.0.2: https://github.com/huangxueling027/DOPO-EP-ML. Original journal articles, publisher supplementary files, and copyrighted figures are not redistributed in this archive.
Files
Steps to reproduce
The dataset contains the processed modeling data and final analysis outputs used in the manuscript. Reproducibility code is available in the DOPO-EP-ML GitHub repository, release v1.0.2. The modeling workflow includes task-specific molecular and structured feature generation, molecule-grouped nested cross-validation, interpretation and applicability-domain analyses, independent literature-based validation, and multi-objective virtual screening. Detailed execution instructions are provided in the repository documentation.
Institutions
- Shanghai University of Engineering ScienceShanghai, Shanghai