Matrices of descriptors can be used to compare physicochemical properties of proteins on their surfaces. Such comparison can be performed introducing similarity indices (SI) defined for example as Euclidean distances between descriptors:

where indices i and j enumerate molecules and indices r and l run over coordinates of 2D grids (columns and rows of descriptors matrices). For a group of proteins, one can construct similarity (or distance) matrices that can be used as an input for various clusterization procedures.