Identification of membrane PAINS via an optimized computational protocol

Membrane Pan-Assay INterference compoundS (PAINS) are a subcategory of molecules that interact with
lipid membranes in a nonspecific way and alter their physicochemical properties [1]. A prompt detection
of these compounds in the drug discovery process is therefore crucial, as it avoids wasting precious time
and resources chasing after false leads. Here we present an optimized umbrella sampling/molecular
dynamics-based computational protocol to identify compounds with varying degrees of membrane PAINS
behavior. We observed that the method was extremely susceptible to fluctuations in membrane
thickness, which we were able to alleviate by changing the US-reference position from the membrane
center to the closest interacting monolayer. The computational performance was further improved by
adjusting the number, strength, and position of the umbrellas. The membrane permeability coefficients
calculated using the inhomogeneous solubility diffusion model were able to accurately assess the
membrane PAINS character of both curcumin and resveratrol [2] but indicated a possible misclassification
of notophagin in a previous work [3].
Acknowledgements: FCT to projects PTDC/BIA-BFS/28419/2017 (B. L. Victor) UIDB/04046/2020−UIDP/04046/2020
(BioISI).
[1] Baell, J.B. et al. (2010) JMC 53(7):2719 doi.org/10.1021/jm901137j
[2] Ingólfsson, H.I. et al. (2014) ACSCB 9(8):1788 doi.org/10.1021/cb500086e
[3] Jesus, A.R. et al. (2017) JMC 60(2), 568-579 doi.org/10.1021/acs.jmedchem.6b01134