A paper on cyclic peptide membrane permeability prediction, first-authored by Masatake Sugita, a researcher in our group, has been published in the open-access journal ACS Omega.
In this study, we developed a machine learning protocol that combines 3D descriptors derived from conformations obtained by molecular dynamics (MD) simulations with 2D descriptors derived from cyclic peptide chemical structures, aiming to improve generalizability while reducing the simulation cost relative to direct MD-based prediction. Across 252 cyclic peptides, the best performance was obtained with XGBoost (correlation coefficient R = 0.77), and the model also showed good accuracy on an external data set not included in training.
Masatake Sugita, Yudai Noso, Jianan Li, Takuya Fujie, Keisuke Yanagisawa, Yutaka Akiyama, "Protocol for Membrane Permeability Prediction of Cyclic Peptides Using Descriptors Obtained from Extended Ensemble Molecular Dynamics Simulations and Chemical Structures", ACS Omega 11, 40628-40644, 2026/6. DOI: 10.1021/acsomega.6c03721
(translated by Claude)