Dataset: MatCreatioNN: Machine Learning Guided Discovery of Photocatalysts for Environmental Applications

MatCreatioNN – Zenodo Archive This Zenodo record contains all datasets and supporting files required to reproduce the machine-learning framework, prediction funnel, and stability/catalysis filters used in the MatCreatioNN workflow. The archive provides complete transparency and reproducibility for the ML-driven screening of Metal–Organic Frameworks (MOFs). Included in this Archive MOF_MachineLearningDatasets/Contains the eight curated MOF datasets used to train the CGCNN models in the multi-stage screening funnel. These datasets cover:• Thermal stability• Water stability• CO₂ adsorption energies• CO₂/H₂O selectivity• Available surface area• Material cost estimation• Sustainability (waste-derived element classification)• Synthesis likelihood These datasets directly correspond to the thirteen trained CGCNN predictors used in the study. SupplementalDatasets/This directory contains several key datasets used in the reproducibility study, including:• MOFSimplify Thermal Stability dataset• MOFSimplify Water Stability dataset• Atomic-scale MOF cost breakdown (element-level industrial cost assignments)• List of sustainable atoms as defined in ref.• 381-material experimental dataset for researchers interested in verifying generalizability of this workflow to other activated-oxygen-photocatalysis (AOP) domains These datasets support training, benchmarking, and external replication of catalytic and materials-property models. Overview MatCreatioNN is a computational pipeline for the generation, evaluation, and property prediction of MOFs.The full workflow integrates:• Crystal Graph Convolutional Neural Networks (CGCNNs)• AI-assisted MOF generation• Quantile-based multi-stage filtering• Stability, adsorption, sustainability, and cost scoring This Zenodo archive provides the datasets necessary to reproduce all machine-learning training runs and validation tests used in the accompanying study. For Code, Trained Models, and Implementation The source code, trained CGCNN models, and supporting scripts used in this project are available on GitHub:https://github.com/SatyaK-0/MatCreatioNN License Unless otherwise noted, data is released under the MIT License or the associated dataset license.