A Generic Model for Mixture Adsorption Property Prediction (MAPP): Toward Efficient Discovery of Metal-Organic Frameworks for CO$_2$ Capture

This repository contains the training data and trained model files used in our work titled "A Generic Model for Mixture Adsorption Property Prediction (MAPP): Toward Efficient Discovery of Metal-Organic Frameworks for CO$_2$ Capture". The goal of this study is to predict CO$_2$/N$_2$ adsorption in MOFs under various conditions (e.g. pressure and gas composition). The training data were curated from GCMC simulation using RASPA 2.0.47, and the model was trained using Pytorch 1.13.1. Version History Version 2.0 (Current): Updated models and extended data Added MAPP-V4 models (ExTransformerV4 architecture) with improved performance Added CGCNN baseline models for comparison Added pure component model for IAST predictions Added extended training data with original absolute loading and excess loading data, and multiple transformation formats (log10/symlog10/arcsinh) Added experimental MOF validation data (IRMOF-1, UiO-66, MIL-53(Al), CALF-20) Version 1.0: Initial release with V3 models (available via version history) Included Files (Version 2.0) Data Files: cifs_graphs_grids.tar.gz: Processed MOFs' CIF, CGCNN graph, and energy grid files (inherited from v1) mof_split_val1000_test1000_seed0_org_csv.zip: Labeled data with multiple adsorption representations, MOF-based split mof_cluster_split_val1_test3_seed0_org_csv.zip: Labeled data with multiple adsorption representations, cluster-based split isotherm_extra_test.tar.gz: Extra test set isotherm data from GCMC simulation processed_data.tar.gz: Processed training data (id_condition_ads_qst_org_all.csv) exp_mof_data.tar.gz: Experimental MOF validation data and simulation data (5 MOFs) MAPP-V4 Models (MOFTransformer-based): model_MAPP_GMOF_v4.tar.gz: MAPP model trained on MOF-based split model_MAPP_GCluster_v4.tar.gz: MAPP model trained on cluster-based split model_MAPPPure.tar.gz: Pure component model for IAST predictions CGCNN Baseline Models: model_CGCNN_GMOF.tar.gz: CGCNN model trained on MOF-based split model_CGCNN_GCluster.tar.gz: CGCNN model trained on cluster-based split