Permeability prediction in the beyond-Rule-of 5 chemical space: Focus on cyclic hexapeptides

Abstract

Cyclic peptides (CPs) are gaining more and more relevance in drug discovery. Since one of their main drawbacks is poor permeability, the discovery of new orally available CP drugs requires computational tools that predict CP permeability in very early drug discovery.In this study we used a literature dataset of 62 cyclic hexapeptides to evaluate the performances of a number of in silico tools based on different computational theory to model and rationalize PAMPA and Caco-2 permeability values. In particular, we submitted the dataset to a) online calculators, b) QSPR strategies, c) a physics-based tool, d) a mixed approach and e) a kinetic method. This latter is an emergent strategy in which a few relevant conformations retrieved from a set of molecular dynamics (MD) simulations by the Markov State Model (MSM) are used to establish the compounds permeability. Both free and commercial software were used. Results were compared with a model based on experimental physicochemical descriptors. All the computational approaches but online calculators performed quite well and show that lipophilicity and not polarity is the main determinant of the investigated event. A second major outcome of the study is that the impact of flexibility on the permeability of the considered dataset cannot be unambiguously assessed. Finally, our comparative analysis, which also included not common applied strategies, allowed a sound evaluation of the pros and cons of the applied computational approaches.