Characterization and Similarity Measurement of Molecules Using Molecular Descriptors

Abstract

When working with molecules as data, you need to describe their structures and properties in a numerical way. And that is what molecular descriptors are used for. Molecules can thus be seen as points in multidimensional space and one can measure similarity and distance between them using various coefficients and metrics and also apply machine learning methods to predict not yet known physico-chemical properties and biological activity of the system. In this article, principles and examples of molecular descriptors, as well as their use to measure similarity and distance between molecules, are shown.