High Structural Error Rates in “Computation-Ready” MOF Databases Discovered by Checking Metal Oxidation States
Andrew J. P. White, Marco Gibaldi, Jake Burner, Rebeca Mayo, Tom K. Woo
Abstract
"Computation-ready" metal-organic framework (MOF) databases provide essential raw data for high-throughput computational screening (HTS) and machine-learning approaches to materials discovery. However, the structural fidelity of these databases remains largely unquantified. We introduce MOSAEC, an algorithm that detects chemically invalid structures based on metal oxidation states. MOSAEC was manually validated against 14,796 MOF structures from the popular CoRE database and found to flag erroneous structures with 96% accuracy. Examination of 14 leading experimental and hypothetical MOF databases containing >1.9 million structures reveals structural error rates exceeding 40% in most cases. Analysis of 8 recent HTS studies which highlighted top-performing candidates shows that 52% of these structures were chemically invalid.