Introduction

SEING streamlines the process of computing fingerprints of molecular systems with a focus on those that explicitly encode spatial coordinates.

Fingerprints (in this context) are numerical representations of chemical structures designed to be invariant under property-perseving operations such as permutation of atoms of the same nature, geometric rotation, etc.

Inspired by similar representations in chemi-informatics, those structural representations were created as alternatives to cartesian coordinates which are not suitable for machine learning studies. The hope is that those fingerprints will open the door for new strucutre-property explorations and the development of improved predictive capability in materials science.

SEING (old French word for “signature”) is created and released to the community as a vehicle to streamline and facilitate the use of those fingerprints in ML applications in the hope of accelerating the use of AI in materials science applications for better understanding and easier discovery of a wide range of new materials.