Rapid classification of bi-heterocyclic compounds by using THz-TDS with combined double-loop learning SVM and kNN

We present a method for rapid recognition of pharmaceutical bi-heterocyclic compounds based on THz spectroscopy (THz-TDS) and a self-optimizing SVM algorithm combined with kNN. Our research focuses on the classification of compounds that have not yet entered into production, but are in the 'in vitro' testing phase (e.g. COX1, COX2 enzyme inhibition, metabolism analysis [1], [2]). The developed method allows error-free verification of known and unknown compounds in a quasi-real time.