QC Singlepoint Dataset: MLPepper RECAP Optimized Fragments v1.1

A single point dataset created by combining the 50k ESP from Simon and Br substituted set from Lily. This was then extended to cover Boron and Silicon by applying the RECAP decomposition scheme in RDKit to the B & Si pubchem set from SPICE2. This dataset now contains additional iodine containing molecules. Each fragment had 5 conformations generated which were optimised locally using an AIMNET2 model trained to wb97m-d3. The aim of the dataset is to provide polarised and gas phase electrostatic properties which can be used to generate ML models for partial charge prediction. Unlike past datasets the wavefunction will not be saved to recompute the ESP instead we recommend building the ESP from the MBIS atomic multipoles which save substantial amount of space. An off equilibrium data set will also be generated to enable conformation dependent prediction of charges. Additional details can be found in the OpenFF QCA Dataset Submission page for this dataset. General Information * Date: 2025-07-01* Class: OpenFF SinglePoint Dataset* Purpose: Single point property calculations for charge models, expanded to include iodine* Name: MLPepper RECAP Optimized Fragments v1.1* Number of unique molecules: 56351* Number of conformers: 75097* Number of conformers (min, mean, max): 1.00, 1.33, 5.00* Molecular weight (min, mean, max): 32.12, 163.20, 701.59* Charges: -4.0, -3.0, -2.0, -1.0, 0.0, 1.0, 2.0, 3.0* Dataset submitter: Jennifer A Clark* Dataset curator: Josh Horton / Charles Adams* Dataset generator: Josh Horton / Charles Adams