A Robotics-Inspired Screening Algorithm for Molecular Caging Prediction

Abstract

We define a molecular caging complex as a pair of molecules, where one molecule (a "host", or a "cage") possesses a cavity that can encapsulate the other molecule (a "guest") and prevent it from escaping. Molecular caging complexes can be useful in applications such as molecular shape sorting, drug delivery, and molecular immobilization in material science, to name just a few. However, the design and computational discovery of new caging complexes is a challenging task, as it is hard to predict whether one molecule can encapsulate another due to the fact that their shapes can be quite complex. In this paper, we propose a computational screening method that predicts whether or not given pairs of molecules form a caging complex. Our method is based on a caging verification algorithm that was designed by our group for applications in robotic manipulation. We test our algorithm on three pairs of molecules that were previously described in a pioneering work on molecular caging complexes, and show that our results are fully consistent with the previously reported ones. Furthermore, we perform a screening experiment on a dataset consisting of 46 hosts and 4 guests, and use our algorithm to predict which pairs are likely to form caging complexes. Our method is computationally efficient and can be integrated in a screening pipeline to complement experimental techniques.