Accurate specification of molecular structures: the case for zero-order bonds and explicit hydrogen counting.

Most data structures used to represent molecular entities for cheminformatics are underspecified for purposes of representing nonorganic chemical species. Two extensions are proposed: allowing bond orders of 0 and adding an atom property to control the number of inferred attached hydrogen atoms. The case for these two extensions is made by demonstrating the effective representation of a number of unconventional bonding types that cannot be effectively represented by data structures currently in common use. A set of enhancements to the industry standard MDL CTfile format is proposed, which includes a backward compatibility mechanism to maximize interpretability by software that has not been updated to make use of the extensions.