Stochastic molecular descriptors for polymers. 1. Modelling the properties of icosahedral viruses with 3D-Markovian negentropies

The application of topologic descriptors to polymer data is considered to be an emerging field. Recently, MARCH-INSIDE was introduced, a topologic approach based on Markov chains. In an attempt to extend its applications to biopolymers, we introduce the propagation entropies of an electrostatic interaction using a protein road map. These negentropies are then used to classify Human Rhinoviruses as binding to the low-density lipoprotein receptor (100%) or intracellular adhesion molecule, with an accuracy of 88.89%. Overall accuracy varies between 91.6 and 100% in training and leave-group-out cross validation. In qualitative terms, the study opens the way to the application of topologic indices in biopolymer modelling.