Structure elucidation in chemoinformatics aims to transform heterogeneous spectral data into molecular graphs. The complexity of the structures often requires manual elucidation which proves to be powerful but very time-consuming and requires the presence of an expert. These limitations have led to the need for automated methods that integrate the ever-evolving elucidation rules into machine-readable processes for predicting results while recording intermediate steps to support their reasoning. We present two complementary rule-based workflows. The additive approach constructs molecular graphs using a strict rule set, while a subtractive approach broadly generates hypotheses and then prunes them with loose constraints. Both start from tabular inputs and build a structured output alongside an auditable chain of intermediate results to support their reasoning. The approaches formalize good reasoning while yielding well-justified candidate sets. In the event they cannot provide the correct result, the output structural consistencies and the tracked information on the intermediate data prove very valuable for experts to continue the work. This clarity of intermediate results and reasoned assembly strongly differentiate these two aproaches from the current automated methods. The approaches thus reduce analysis time, highlight shortcomings, support instruction, and lay the foundation for future CASE systems.
|
