Development of image similarity strategy based on targeted filtration for non-targeted HS-SPME/GC × GC fingerprints of volatile oils from Chinese patent medicines: A case of Chaihu Shugan Wan

Abstract

To evaluate Chinese patent medicines more comprehensively, image similarity strategy based on targeted filtration was developed for non-targeted two-dimensional (2D) chromatographic fingerprints of volatile oil fraction in this paper. At first, volatiles and semi-volatiles were enriched from Chaihu Shugan Wan (CSW) by head space- solid phase micro-extraction (HS-SPME), and analyzed by comprehensive two-dimensional gas chromatography (HS-SPME-GC × GC). In succession, the inter-group and intra-group differences were distinguished by multi-component spectral correlative chromatography (MSCC). Facing the challenge of data complexity, some 50 contours were selected to indicate various medicinal materials through zone- specific ion filtration in 2D fingerprints. Among them, the known compounds were proven to be effective in “soothing liver-qi stagnation” through PubChem and other bioinformatic tools. Thirdly, Speeded Up Robust Features- Fast Library for Approximate Nearest Neighbors (SURF-FLANN) algorithm was utilized to estimate the pairwise similarities for the simplified fingerprints. In this procedure, the methodological examination was further made through the simulated images, including comparison with other algorithms, consideration for image complexity, and calculation of positives / negatives. At last, the similarities of six sample fingerprints were scored successfully. In summary, image similarity strategy based on zone- specific ion filtration can provide a viable solution for rapidly evaluating 2D fingerprints of extract fractions from Chinese patent medicines.