CycPeptMPDB-4D: Multi-Solvent Conformational Ensembles for Predicting Cyclic Peptide Permeability

CycPeptMPDB-4D: A Conformational Dynamics Dataset of Cyclic Peptides for Membrane Permeability Prediction CycPeptMPDB-4D is a large-scale structural dynamics dataset featuring atomistic molecular dynamics (MD) trajectories and 3D conformational ensembles for 5,160 structurally diverse cyclic peptides. It extends CycPeptMPDB by adding MD-derived conformations and physics-based molecular descriptors. This resource is designed to support the development of 3D and 4D (trajectory- or ensemble-based) deep learning models for predicting membrane permeability, bridging the gap between static 2D representations and the physical mechanisms of peptide–membrane interaction. A key feature of this dataset is the inclusion of simulations in both explicit water and hexane environments. This dual-solvent approach captures the "chameleon-like" behavior of cyclic peptides—their ability to adopt distinct conformational states to maximize stability in aqueous environments and minimize polar exposure in hydrophobic membrane cores. More details can be found here: (Link to be added upon publication) Dataset Structure The data is organized into solvent-specific directories containing structural information and a central metadata file: CycPeptMPDB-4D/├── Water/ 5,160 peptides│ ├── Trajectories/ *.pdb (100 frames per peptide)│ ├── Structures/ Representative conformations from clustering│ └── Logs/ Clustering analysis logs├── Hexane/ 5,160 peptides│ ├── Trajectories/ *.pdb (100 frames per peptide)│ ├── Structures/ Representative conformations from clustering│ └── Logs/ Clustering analysis logs├── CHCl3/ 6 peptides└── CycPeptMPDB-4D.csv metadata & molecular descriptors Update History 9-July-2026: Fixed stereochemical inconsistencies (D/L configuration errors), mainly in peptides with unnatural amino acids, BHF and TNH. Added a few more columns to CycPeptMPDB-4D.csv, and converted all units from nm to Å.