Modeling Molecularly Imprinted Nanoparticles with LNKD: A Resource Efficient Algorithm for Polymer Crosslinking

Chemical crosslinking has been widely used to modify the physical and chemical properties of materials. Molecularly imprinted polymers (MIPs) are a sub-class of crosslinked polymers. Their designable binding sites make them highly useful in a broad range of chemical and biological applications. Computational efforts to model, characterize, and design crosslinked polymers are limited in part due to challenges in obtaining their matrix-like and probabilistic structures experimentally. Computational prediction of polymer crosslinking is resource-intensive and underexplored. Here, we propose LNKD (Linking Nodes in KD-trees), a resource-efficient algorithm for predicting pairs of reactive atoms in pre-crosslinked 3D structures of monomers that applies not only to the modeling of MIPs, but also chemical crosslinking in other materials. LNKD performs a spatial query around all reactive atoms in a structure and uses a crosslinking probability function to predict pairs of atoms most likely to form crosslinks. Additionally, we introduce a protocol for modeling molecularly imprinted nanoparticles (MINPs), a type of MIP, combining molecular dynamics simulations with LNKD. We validate the method by its accurate modeling of MINPs in their binding properties in comparison to experimental results. For the MINPs tested, LNKD found crosslinking pairs for 88-95% of the 780 total reactive atoms in approximately three seconds on a laptop and docking results reproduce experimental trends in ligand binding.