A novel digital identification strategy for licorice raw materials: TCM-ID with common/exclusive ion matrices and hierarchical matching

Licorice is one of the most widely used medicinal plants on Earth, appearing in thousands of traditional prescriptions, cough syrups, and modern pharmaceutical formulations. Yet the dried roots and rhizomes sold under the name Gan Cao, or Glycyrrhizae Radix et Rhizoma, can come from several different botanical species, and telling them apart has long been a stubborn problem for regulators. Now a team at China’s National Institutes for Food and Drug Control has unveiled a digital fingerprinting strategy that assigns each licorice species a kind of mass spectrometric identity card, allowing a computer to authenticate the plant’s origin with remarkable precision. The approach, described in Frontiers in Chemistry, combines high-resolution liquid chromatography-mass spectrometry with a concept the researchers call TCM-ID, or traditional Chinese medicine digital identity.

The two species at the heart of the study are Glycyrrhiza uralensis Fisch. and Glycyrrhiza inflata Bat., which together account for more than 85 percent of the medicinal licorice circulating in the Chinese market. Both are official botanical origins of Gan Cao, but their chemical profiles differ in ways that matter for efficacy and clinical safety. Traditional authentication relies on morphological and microscopic features such as cork cell arrangement, fiber bundle compactness, and the abundance of calcium oxalate crystals. These characteristics overlap heavily between the two species, and judgments depend on the subjective experience of the identifier. Molecular methods based on chloroplast DNA markers can discriminate the species, but they typically require fresh samples and are too cumbersome for the high-throughput demands of commercial quality control.

The researchers’ solution starts with a deceptively simple observation: even when licorice samples differ in geographical source, growth years, or harvest batch, material from the same species inevitably contains the same core chemical components. The team analyzed 35 batches of G. uralensis and 35 batches of G. inflata, all botanically verified against the 2025 Chinese Pharmacopoeia, using a Waters QTOF high-resolution mass spectrometer. Samples were ground, sieved, and extracted with 75 percent methanol under ultrasonication, then separated on a UHPLC C18 column with a 45-minute gradient elution. Detection ran in both positive and negative electrospray ionization modes, scanning masses from 50 to 1200 daltons. The raw data were converted into a three-dimensional matrix of retention time, mass-to-charge ratio, and ion intensity, yielding between roughly 4,000 and 9,500 ion features per batch.

From this torrent of data, the researchers constructed two complementary matrices. The first, the common ion matrix, or CIM, captures the twenty most intense ions that recur consistently across every batch of a species, regardless of origin, growth year, or storage conditions. These ions represent the stable chemical backbone shared by all genuine samples of that species. The second, the exclusive ion matrix, or EIM, is built by removing the ions that the two species share and retaining the twenty most intense species-specific features. The distinction matters because the two licorice species are chemically close cousins: their total ion chromatograms look strikingly similar, and both contain compounds such as liquiritigenin and liquiritin at nearly identical retention times and masses.

That chemical kinship became obvious in the first round of matching. When test samples were compared against the common ion matrices, every G. uralensis sample scored a matching confidence of at least 90 percent against its own species matrix, but it also scored 60 to 90 percent against the G. inflata matrix. The confidence increment rate, a novel parameter the team devised to express how much better a sample matches one species than the other, ranged only from 0.06 to 0.58. In other words, the common ion matrix alone could provide a preliminary screening but could not deliver unambiguous discrimination. The researchers therefore designed a hierarchical matching strategy in which the CIM serves as the first gate, requiring a matching confidence of at least 90 percent, before the sample proceeds to the decisive second stage.

The second stage, based on the exclusive ion matrices, proved dramatically more selective. G. uralensis samples scored 80 to 100 percent against their own exclusive matrix while dropping to zero or 5 percent against the G. inflata matrix, with confidence increment rates exceeding 17. G. inflata samples scored at least 70 percent against their own exclusive matrix and no more than 10 percent against the foreign one. Through cutoff optimization across an external validation set of 30 batches, the team established a dual criterion: a sample is authenticated only if its matching confidence reaches at least 70 percent and its confidence increment rate reaches at least 6.00, applied as a logical AND. Receiver operating characteristic analysis of this combined model returned an area under the curve of 1.00, with 100 percent sensitivity and 100 percent specificity and no false positives or false negatives.

The method’s practical power was demonstrated on ten batches of commercial licorice purchased blind from the market. Eight batches were authenticated as G. uralensis and two as G. inflata, with the market samples showing matching confidences of at least 70 percent against their own exclusive matrices and confidence increment rates of 12 or higher. Method validation confirmed the analytical foundation: mass errors stayed within 5 parts per million, precision relative standard deviations remained below 0.14 percent, and repeatability and stability deviations stayed below 0.70 percent over 48 hours. The blank solvent produced no significant interference, confirming the method’s specificity.

Chemometric analysis added a second layer of evidence. Using UPLC-QTOF-MS data from 20 batches of each species, the researchers applied principal component analysis and orthogonal partial least squares discriminant analysis, which cleanly separated the two species with an R2Y of 0.989 and a Q2 of 0.992. A permutation test and CV-ANOVA confirmed the model was valid and not overfitted. Screening markers with a variable importance in projection threshold of 10.0 revealed seven differential chemical markers. For G. inflata, the markers were licochalcone A, licochalcone B, echinatin, and glabrol, with VIP values up to 25.01. For G. uralensis, the markers were licoricidin, licoricone, and glycycoumarin. Ion distribution profiles showed these compounds were prominent in one species and nearly absent in the other, consistent with earlier reports in the literature.

The deeper significance of the work lies in its conceptual shift. TCM-ID moves authentication away from the structural elucidation of individual marker compounds and toward the digital quantitative characterization of the entire chemical profile, making full use of massive datasets that include thousands of unidentified components. Because the common ion matrix absorbs the natural variability caused by cultivation region, growth years, and harvest batch, while the exclusive ion matrix captures the interspecific differences, the combined criteria avoid the limitations of single-index identification. The researchers note that the same strategy has already been applied to angelica, saposhnikovia, and bile powder, suggesting a generalizable framework for botanical origin traceability across many traditional medicines.

Limitations remain. The study focused only on G. uralensis and G. inflata, excluding Glycyrrhiza glabra, the third official licorice species, so the current cutoff values apply only to the two species examined. The team plans to add G. glabra specimens and multi-batch data to optimize the matrices and thresholds for comprehensive three-species identification. Still, the methodological framework is highly replicable, and the authors argue that TCM-ID offers a promising route for origin traceability and quality evaluation of herbal medicines. As traditional Chinese medicine enters its digital and intelligent era, the licorice identity card may become a template for how regulators authenticate the botanical origins of the world’s herbal pharmacopoeia, one mass spectrum at a time.

Subject of Research: A novel digital identification strategy for licorice raw materials: TCM-ID with common/exclusive ion matrices and hierarchical matching

Article Title: A novel digital identification strategy for licorice raw materials: TCM-ID with common/exclusive ion matrices and hierarchical matching

Article References: Wang, X., Wu, Y., Wan, L., Zhang, J., He, F., Jing, W., Chen, J., Cheng, X., & Lin, Y. (2026). A novel digital identification strategy for licorice raw materials: TCM-ID with common/exclusive ion matrices and hierarchical matching. Frontiers in Chemistry, 14, Article 1918761. https://doi.org/10.3389/fchem.2026.1918761

Image Credits: AI Generated

DOI: 10.3389/fchem.2026.1918761

Keywords: novel, digital, identification, strategy, licorice, materials, TCM-ID, common, exclusive, matrices, hierarchical, matching