Sweet Flavor as a Strong Discriminator of Tonifying versus Eliminating Action

Abstract This study does not take chemical constituents as its starting point; instead, it begins with the measurable labels of medicinal properties themselves and examines the relationship among the four natures, the five flavors, and tonifying versus eliminating action. Habitat, as a more upstream antecedent framework, is examined independently in a companion line of work and is not used here. The nature of the measurement must first be defined: the four natures, five flavors, and functional classifications all come from the same textbook–Chinese Pharmacopoeia (2020 edition) label system, in which the sweet flavor is defined as that which can tonify, can moderate, and can harmonize. What this study measures is therefore the internal consistency and out-of-fold reproducibility within that label system, not causal relationships in the natural world. The analysis is restricted to the binary contrast of tonifying deficiency versus eliminating pathogenic factors: the primary analysis frame has n = 366 (80 tonifying-deficiency records and 286 from the twelve eliminative classes), and the focused frame has n = 172 (80 tonifying-deficiency and 92 heat-clearing records). Given the complete separation observed in the focused 172 frame—24 of 24 sweet-containing records on the hot side and 2 of 52 non-sweet records on the cold side—we use Firth's penalized likelihood with a Jeffreys prior, supplemented by Fisher's exact test. Results: on the cold side, where controls are available, the tonifying-versus-eliminating contrast between sweet-containing and non-sweet records gives a Fisher odds ratio of 17.19 (p = 3.9 × 10⁻⁶); the neutral group gives 25.0 (p = 0.001), with the two directions agreeing in sign and comparable in magnitude. On the hot side, 29 of 30 records tonify deficiency and only one is heat-clearing (and that record is non-sweet), so there is only a single eliminating record, insufficient as a control, and discriminative power on that side cannot be tested. The 36 neutral records, of which 28 are sweet-containing and 27 tonify deficiency while 9 do not, are reported separately as the neutral group. Under five-fold family-grouped GroupKFold, AUC is 0.644 for the four natures alone, 0.778 for the five flavors alone, and 0.789 for the combined model; the increment of the combined model over the five flavors alone is only 0.010 (95% CI -0.027 to 0.05). Under the symmetric scale (a comparable calibration), the increment is 0.025 and the standalone discriminative power of the four natures falls to 0.599; the third calibration runs in the opposite direction to the baseline, and its value of 0.037 serves only as an upper-bound reference. None of the three reaches 0.06, which is used here only as a magnitude reference. The increment is an exploratory analysis, registered in research protocol v1.7 as the exploratory addendum H10. The single binary indicator of sweet presence has AUC 0.762, differing from the full five flavors by 0.016 (CI -0.028 to 0.081). When the frame is widened from 366 to 391 (adding 25 digestion-promoting records) and to 427 (all non-tonifying records), the sweet-flavor odds ratios remain 13.08 and 11.95 (both p < 10⁻¹⁵), while the proportion of tonifying records among sweet-containing medicinals falls from 0.428 to 0.340. Under the pre-registered threshold (upper bound of the difference <= 0.06), the current upper bound of 0.081 does not meet the threshold for the carrier designation. We therefore do not use the stronger formulation that the sweet flavor alone accounts for tonifying action; instead we characterize the sweet flavor as a strong discriminator, while the bitter and pungent flavors may still retain some increment.

Authors

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11712123/v1
Primary Topic
Traditional Chinese Medicine Studies
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Sweet Flavor as a Strong Discriminator of Tonifying versus Eliminating Action

Guohua Yang
2 links
Research Square
Traditional Chinese Medicine Studies
preprint

Sweet Flavor as a Strong Discriminator of Tonifying versus Eliminating Action

Guohua Yang
preprint en

Abstract

Abstract This study does not take chemical constituents as its starting point; instead, it begins with the measurable labels of medicinal properties themselves and examines the relationship among the four natures, the five flavors, and tonifying versus eliminating action. Habitat, as a more upstream antecedent framework, is examined independently in a companion line of work and is not used here. The nature of the measurement must first be defined: the four natures, five flavors, and functional classifications all come from the same textbook–Chinese Pharmacopoeia (2020 edition) label system, in which the sweet flavor is defined as that which can tonify, can moderate, and can harmonize. What this study measures is therefore the internal consistency and out-of-fold reproducibility within that label system, not causal relationships in the natural world. The analysis is restricted to the binary contrast of tonifying deficiency versus eliminating pathogenic factors: the primary analysis frame has n = 366 (80 tonifying-deficiency records and 286 from the twelve eliminative classes), and the focused frame has n = 172 (80 tonifying-deficiency and 92 heat-clearing records). Given the complete separation observed in the focused 172 frame—24 of 24 sweet-containing records on the hot side and 2 of 52 non-sweet records on the cold side—we use Firth's penalized likelihood with a Jeffreys prior, supplemented by Fisher's exact test. Results: on the cold side, where controls are available, the tonifying-versus-eliminating contrast between sweet-containing and non-sweet records gives a Fisher odds ratio of 17.19 (p = 3.9 × 10⁻⁶); the neutral group gives 25.0 (p = 0.001), with the two directions agreeing in sign and comparable in magnitude. On the hot side, 29 of 30 records tonify deficiency and only one is heat-clearing (and that record is non-sweet), so there is only a single eliminating record, insufficient as a control, and discriminative power on that side cannot be tested. The 36 neutral records, of which 28 are sweet-containing and 27 tonify deficiency while 9 do not, are reported separately as the neutral group. Under five-fold family-grouped GroupKFold, AUC is 0.644 for the four natures alone, 0.778 for the five flavors alone, and 0.789 for the combined model; the increment of the combined model over the five flavors alone is only 0.010 (95% CI -0.027 to 0.05). Under the symmetric scale (a comparable calibration), the increment is 0.025 and the standalone discriminative power of the four natures falls to 0.599; the third calibration runs in the opposite direction to the baseline, and its value of 0.037 serves only as an upper-bound reference. None of the three reaches 0.06, which is used here only as a magnitude reference. The increment is an exploratory analysis, registered in research protocol v1.7 as the exploratory addendum H10. The single binary indicator of sweet presence has AUC 0.762, differing from the full five flavors by 0.016 (CI -0.028 to 0.081). When the frame is widened from 366 to 391 (adding 25 digestion-promoting records) and to 427 (all non-tonifying records), the sweet-flavor odds ratios remain 13.08 and 11.95 (both p < 10⁻¹⁵), while the proportion of tonifying records among sweet-containing medicinals falls from 0.428 to 0.340. Under the pre-registered threshold (upper bound of the difference <= 0.06), the current upper bound of 0.081 does not meet the threshold for the carrier designation. We therefore do not use the stronger formulation that the sweet flavor alone accounts for tonifying action; instead we characterize the sweet flavor as a strong discriminator, while the bitter and pungent flavors may still retain some increment.

Research Square
Traditional Chinese Medicine Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon