An ab initio softness metric to measure the similarity between all pairs of amino acids

The search for quantitative index of similarity between molecular moeties is important for its applications in pharmacology. Similarity is also an important concept in computational biology to measure the exchangeability of an amino acid by another in a protein sequence. In the present work, a distance between two molecules based on local and global softnesses of their fragments is defined. The method proposed is general and could be applied to any molecular library. It is first applied to compute the distance between the 190 pairs of different amino acids in their neutral states. Two amino acids belonging to the one of the biochemical class (aliphatic, sulfur-containing, acidic, ...) are found to be at a smaller distance than two amino acids belonging to two different classes. A first attempt to relate this ab initio distance to the substitution matrix used in computational biology is presented.