A critical evaluation of automatic atom mapping algorithms and tools

Abstract

The identification of the atoms which change their position in chemical reactions is an important knowledge within the field of Metabolic Engineering. This can lead to new advances at different levels from the reconstruction of metabolic networks to the classification of chemical reactions, through the identification of the atomic changes inside a reaction. The Atom Mapping approach was initially developed in the 1960s, but recently suffered important advances, being used in diverse biological and biotechnological studies. The main methodologies used for atom mapping are the Maximum Common Substructure and the Linear Optimization methods, which both require computational know-how and powerful resources to run the underlying tools. In this work, we assessed a number of previously implemented atom mapping frameworks, and built a framework able of managing the different data inputs and outputs, as well as the mapping process provided by each of these third-party tools. We evaluated the admissibility of the calculated atom maps from different algorithms, also assessing if with different approaches we were capable of returning equivalent atom maps for the same chemical reaction.