Computational modeling has been widely used for decades to predict bioactivities of small molecules in the drug discovery procedure. However, the applicability of computational modeling to deal with nanomaterials (NMs) is limited due to the complexity of NM structures. The lack of suitable descriptors, which can correctly quantify nanostructures and represent the nanostructural diversity, prevents the development of predictive computational models for NMs. To address this challenge, a computational workflow was established in this study to 1) virtually construct diverse gold nanoparticle (GNP) libraries; 2) develop predictive Quantitative Nanostructure Activity Relationship (QNAR) models based on newly designed nanodescriptors; and 3) validate the universal applicability of these new nanodescriptors. The core of this modeling procedure is a novel algorithm to employ geometrical chemical descriptors to quantify nanostructures based on constructed virtual gold nanoparticles (vGNPs). The applicability, validity and feasibility of this method were proofed by modeling six in-house GNP datasets and one external dataset, which consist of different nanostructures and were experimentally tested for various nano-bioactivities (i.e., GNP-enzyme interactions, nano cellular uptake and nano-induced reactive oxygen species (ROS)) and physicochemical properties (i.e., logP and zeta potential), respectively. Associated vGNP libraries were first developed based on these GNP datasets. After generating nanodescriptors from all vGNPs, QNAR models were constructed using various machine learning approaches for all the modeling sets. The high predictabilities of all developed QNAR models suggest this workflow can be used as a universal tool for nanomaterial modeling and virtual screening purpose.

Novel nanodescriptors applied in QNAR: Combination of virtual nanomaterial library and geometrical structure of nanomaterial

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.