Molecular docking was utilized to confirm essential targets identified through network pharmacology and to predict the binding locations and affinities between ligands and protein targets [12, 13]. Small-molecule ligands were acquired from the TCMSP or PubChem databases ⁸ in MOL2 format. Protein crystal structures were acquired from the RCSB Protein Data Bank ⁹ in PDB format [14]. Water molecules and existing ligands were eliminated via PyMOL, and the proteins were prepared with AutoDock Tools 1.5.7, which included hydrogenation and conversion to PDBQT format. Ligands were also hydrogenated and ionized prior to docking.

Docking was performed via AutoDock Vina, which generates 20 conformations per ligand‒target pair [15]. The best binding conformation (lowest binding energy) was selected for further analysis. A heatmap of binding affinities was generated via Python 3.13.

Molecular dynamics (MD) simulations were conducted to evaluate the dynamic interactions and binding stability between proteins and ligands. Based on the docking results, three foremost ligand–protein pairs were identified: protein ID:5KIR with naringenin (MOL004328), protein ID:5KIR with divaricatol (MOL011740), and protein ID:8JOW with kaempferol (MOL000422). In this study, GROMACS 2022.5 was used for molecular dynamics simulation [12, 16, 17]. The ambient pH was established at 7.35, the CHARMM36 force field was employed for protein topology, and the ligand topology was produced via the CGenFF service. The system was solvated with TIP3P water, and ions (Na⁺, H⁺) were added to neutralize the system. The simulation steps are as follows:

(1) Energy minimization: First, proteins were constrained, and energy minimization of ligands was performed via the steepest descent algorithm (step size = 0.01, 5000 cycles).

The simulation stability was evaluated by the root mean square deviation (RMSD), the radius of gyration (Rg), the root mean square fluctuation (RMSF), and hydrogen bond analyses. Gibbs free energy landscapes for each complex were calculated and visualized via the DuIvyTools Python library [18].

Senna alexandrina Mill. (Guangxi, batch no.2311300022) was used to induce diarrhea. The composition of Tongxie Yaofang included Citri Reticulatae Pericarpium (Hunan, batch no. 240701), Saposhnikoviae divaricata (Hebei, batch no.240502), Paeoniae Radix Alba (Anhui, batch no.SN24082701), and Atractylodis Macrocephalae (Zhejiang, batch no.SX24092302). All decoction pieces were purchased from Jiuzhitang Pharmacy, Changsha, Hunan Province, China.

A total of 100 g of Senna alexandrina Mill. were immersed in boiling water for 20 min, filtered through sterile gauze, and the filtrate was collected as the first extract. The residue was subjected to a second extraction under the same conditions, and the filtrates were combined. The combined solution was concentrated at 80 °C using a rotary evaporator to obtain an aqueous extract at 100% concentration (1 g/mL), which was stored at 4 °C until use.

Herbs including stir-fried Atractylodis Macrocephalae (8 g), Paeoniae Radix Alba (12 g), Citri Reticulatae Pericarpium (9 g), and Saposhnikoviae Radix (6 g) were soaked in water for 15–20 min. The mixture was boiled over high heat and simmered for 30 min, twice in succession. The two extracts were filtered through sterile gauze, combined, and concentrated to a final crude drug concentration of 0.25 g/mL. The decoction was stored at 4 °C until use [7].

Mouse total bile acid ELISA kit (batch no. JM-11614M2, Jiangsu Jingmei Biological Technology Co., Ltd., China); mouse corticotropin-releasing hormone ELISA kit (batch no. JM-02792M2, Jiangsu Jingmei Biological Technology Co., Ltd., China); mouse motilin ELISA kit (batch no. JM-02775M2, Jiangsu Jingmei Biological Technology Co., Ltd., China).

Thirty SPF-grade, 4-week-old female Kunming mice, each weighing 20 ± 2 g, were acquired from Hunan Slake Jingda Experimental Animal Co., Ltd. [19]. All animals were maintained at the Experimental Animal Center of Hunan University of Chinese Medicine [SCXK (Xiang) 2019--0009] under controlled conditions: temperature 23–25 °C, relative humidity 47–53%, and a standard 12-h light/dark cycle (light from 07:00 to 19:00; dark from 19:00 to 07:00). The Experimental Animal Ethics Committee of Hunan University of Chinese Medicine examined and approved the experimental protocol [Ethics Approval Number: HNUCM21-2409-14].

The mice were fed a clean, pollutant-free standard diet manufactured by Beijing Huafukang Biotechnology Co., Ltd. and supplied by the Experimental Animal Center. The feed certification number is (2024) 06076. The detailed composition and proportions of the diet are shown in Table 1.

Modeling phase: Following 5 days of adaptive feeding, 30 mice were randomly allocated into two groups: a normal control (MC) group (n = 10) and a model (MM) group (n = 20). The modeling procedure was adapted on the basis of a previously published method [20]. All the mice were fasted for 12 h before each gavage. Each morning at 08:30, the mice in the MM group were orally supplied 0.35 mL of Senna alexandrina Mill. to produce spleen deficiency diarrhea, while the MC group received an equivalent dose of distilled water. At 15:00, the mice were subjected to restraint stress by placement in centrifuge tubes to restrict limb movement, while tail-clamping stimulation with long hemostatic clips was applied for 1 h. The procedure involved 15-min intermittent rest periods and was repeated four times. During the modeling phase, mice in both the MM and TX groups were subjected to fasting and water deprivation for seven consecutive days, whereas mice in the MC group had free access to food and water.

TCM treatment phase: Following successful model establishment, the mice in the MM group were randomly assigned to one of two subgroups: the Tongxie Yaofang (TX) group (n = 10) and the MM model control group (n = 10). According to the dose conversion method outlined in the Research Methods in Traditional Chinese Medicine Pharmacology [21, 22], the equivalent dose of the Tongxie Yaofang decoction for mice was determined to be 0.35 mL per administration, twice daily, for 3 consecutive days. The MC and MM groups received an equal volume of distilled water via gavage on the same schedule. The mice were fasted and not given water for 12 h before the operation. The treatment protocol is illustrated in Figure 1.

The clinical characteristics of IBS-D and relevant literature have been used to guide syndrome differentiation, with a focus on changes in fecal properties, aggressive or irritable behaviors, hyperactivity, dull coats, and decreased food and water intake. Mice treated with Tongxie Yaofang showed improvements in mood, coat condition, and fecal characteristics.

Statistical analyses were conducted using SPSS 25.0 software. Quantitative data that adhere to a normal distribution are represented as the mean ± standard deviation (x̅ ± SD). For comparisons between two groups exhibiting normally distributed data, the t-test was employed; conversely, if normality was not satisfied, the Mann–Whitney U test was utilized. One-way analysis of variance (ANOVA) was employed for comparisons among several groups, contingent upon the data meeting the criteria of normality and homogeneity of variance, followed by the least significant difference (LSD) post hoc test. If these assumptions were unmet, the Kruskal–Wallis test was performed. Categorical data were represented as percentages (%) and analyzed via the chi-square test. A significance level of α = 0.05 was established, with p < 0.05 deemed statistically significant.

A total of 40 active ingredients in the four herbs of Tongxie Yaofang were identified through preliminary analysis (see Supplementary Table S1). By screening the TCMSP database, 145 compound-related targets were obtained. Additionally, 4,643 disease-related targets of IBS-D were identified via the use of the GeneCards and OMIM databases. As illustrated in Figure 2A, the Venn diagram displays the overlapping targets between the compounds and the disease, indicating that 28 of the 40 active ingredients (corresponding to 120 targets) were closely associated with the treatment of IBS-D.

To elucidate the multicomponent, multitarget synergistic effects of Tongxie Yaofang in the treatment of IBS-D and to further explore its underlying molecular mechanisms, a compound–target interaction network was constructed. As shown in Figure 2B, the network comprises 148 nodes and 354 edges. In this context, node size is positively correlated with node degree, reflecting the relative importance of a component or target within the network. Therefore, the components with the highest degrees—BS4 (kaempferol, degree = 52), FF11 (wogonin, degree = 40), CP4 (nobiletin, degree = 34), CP1 (naringenin, degree = 32), and B1 (beta-sitosterol, degree = 26)—were identified as the key active compounds in the treatment of IBS-D.

Figure 2C illustrates that protein–protein interaction (PPI) data sourced from the STRING database ² were imported into Cytoscape 3.9.1 to construct the PPI network of Tongxie Yaofang. The five targets with the highest degree values—IL-16 (degree = 85), TNF (degree = 83), AKT1 (degree = 82), TP53 (degree = 77), and PTGS2 (degree = 72)—were identified as core targets potentially mediating the therapeutic effects of Tongxie Yaofang in IBS-D patients, as illustrated in Figure 2D.

GO enrichment analysis was performed to evaluate the overrepresentation of GO terms within the identified gene set, thereby elucidating the shared biological characteristics of these genes in terms of biological processes (BP), cellular components (CC), and molecular functions (MF), as illustrated in Figure 3A. Based on the criteria of p < 0.05 and a false discovery rate (FDR) < 0.05, a total of 216 biological processes, 38 cellular components, and 53 molecular functions were significantly enriched. As shown in Figures 3B–D, the top 20 significantly enriched terms indicated that Tongxie Yaofang may exert therapeutic effects by modulating pathways such as response to xenobiotic stimulus, positive regulation of apoptotic process, caveola, cytosol, enzyme binding, and identical protein binding.

Furthermore, KEGG pathway analysis (p < 0.05, FDR <0.05) revealed 166 enriched signaling pathways, suggesting that the therapeutic action of Tongxie Yaofang in IBS-D patients is closely associated with the AGE-RAGE, TNF, and IL-17 signaling pathways, as presented in Figure 3E. The functional categorization of these KEGG pathways revealed five major classifications: metabolism, environmental information processing, cellular processes, organismal systems, and human diseases. Among these pathways, those related to lipid metabolism, signal transduction, cell growth and death, the endocrine system, and cancer were highly enriched (Figure 3F).

Ultimately, to confirm the connection between the active molecules of Tongxie Yaofang and IBS-D-related targets, a multi-molecular docking study was conducted, and the resultant binding energies were displayed. The top three compound–target complexes exhibiting the most stable binding affinities were further subjected to molecular dynamics simulation, as shown in Figure 3G.

To elucidate the potential molecular mechanisms underlying the therapeutic effects of Tongxie Yaofang on IBS-D, molecular docking was performed using the protein targets 5KIR and 8JOW, along with their respective active ligands: divaricatol (MOL011740), naringenin (MOL004328), and kaempferol (MOL000422). Binding energy, a critical parameter in molecular docking analysis, reflects the stability and affinity of the interaction between a ligand and its target protein. Generally, a more negative binding energy value indicates a stronger and more stable interaction [31].

The binding energy of 5KIR and divaricatol (MOL011740) was calculated to be −9.70 kcal/mol, stabilized by multiple interactions, including van der Waals forces, hydrogen bonds, carbon‒hydrogen bonds, and π‒sigma, alkyl, and π‒alkyl interactions. Five conventional hydrogen bonds were formed with the amino acid residues ARG, GLN, HIS, GLY, and TYR, with corresponding bond lengths of 2.32 Å, 2.54 Å, 1.92 Å, 2.75 Å, and 2.93 Å, respectively.

Similarly, 5KIR and naringenin (MOL004328) exhibited a binding energy of −9.40 kcal/mol, which was supported by van der Waals forces, hydrogen bonds, carbon‒hydrogen bonds, and alkyl and π‒alkyl interactions. Two conventional hydrogen bonds involving residues TYR and GLY, with bond lengths of 2.57 Å and 2.40 Å, respectively, were identified.

The docking of 8JOW and kaempferol (MOL000422) yielded a binding energy of −9.60 kcal/mol, which was also stabilized by van der Waals forces, hydrogen bonds, carbon‒hydrogen bonds, amide‒π stacking, and alkyl and π‒alkyl interactions. Four conventional hydrogen bonds were formed with GLN, TYR, THR, and SER, with bond lengths of 2.93 Å, 2.01 Å, 2.33 Å, and 2.83 Å, respectively.

These results, illustrated in Figure 4, demonstrate that the active compounds of Tongxie Yaofang exhibit strong and stable binding affinities with key IBS-D-associated targets, suggesting a robust pharmacodynamic basis for their therapeutic potential.

As depicted in Figures 6A–C, the mice in the MC group presented a healthy appearance: an alert posture, glossy smooth fur, well‐formed black feces, a clean perianal area, and dry bedding. In contrast, the MM group (Figures 6D–F) presented clinical signs of illness, including hunching, irritability, a tendency to cluster and burrow, dull fur, yellowish loose and sticky stools, severe perianal soiling, and damp bedding. Following the administration of Tongxie Yaofang (Figures 6G–I), the treated mice showed marked improvement: hunching and excessive jumping were alleviated, appetite increased, clustering behavior decreased, fecal color reverted to black, and stool consistency normalized from watery to formed and dry.

As illustrated in Figures 6J,K, fecal samples from MM group mice appeared yellowish, and when they were gently blotted with filter paper, the resulting smear was notably more diffuse than that observed in the MC and TX groups. These findings indicate that Tongxie Yaofang treatment partially restores normal stool consistency and reduces perianal contamination in the liver hyperactivity with spleen deficiency pattern diarrhea model.

Figures 7A,B illustrate that, throughout the modeling period, both food and water intake in the MM and TX groups were inferior to that of the MC group. Following the initiation of Tongxie Yaofang treatment on day 8, water intake in the TX group increased significantly, whereas that in both the MC and MM groups gradually increased. As shown in Figure 7C, body weight increased across all groups. However, after Senna alexandrina Mill. administration, the baseline weights of the MM and TX groups decreased. By day 3 of treatment, the weight difference between MC and MM mice reached significance (p < 0.01). Postmodeling, MC mice weighed significantly more than MM (p < 0.01) and TX (p < 0.05) mice did. From day 8, mice in the TX group received intragastric administration of Tongxie Yaofang, which resulted in a marked increase in body weight. After treatment ended on day 10, the body weight of MC mice was significantly different from that of MM mice (p < 0.05), whereas no significant difference was observed between the TX and MM groups. The defecation frequency (Figure 7D) was similar across all groups at baseline but increased progressively in the MM group during modeling. Following treatment, the defecation frequency of the TX group returned to levels comparable with those of the MC group. Fecal moisture content—a key indicator of diarrhea — was unchanged at baseline (Figure 7E). Compared with the MC group, the MM and TX groups presented significantly greater moisture contents (p < 0.01) throughout the model. After 3 days of treatment, the difference between MC and TX diminished (p < 0.05), and MM versus TX also differed (p < 0.05), confirming the successful induction of diarrhea, in which Tongxie Yaofang improved the diet, water intake, body weight loss, increased defecation, and elevated fecal moisture.

As shown in Figure 8A, the spleen index tended to increase following Senna alexandrina Mill.- and restraint stress modeling but did not differ significantly from that of MC mice (p > 0.05). After 3 days of Tongxie Yaofang treatment, the spleen index in the TX group decreased slightly. Figure 8B shows that the liver index remained unchanged in both the MC and MM groups post-modeling (p > 0.05), whereas the TX group displayed a modest upwards trend after treatment. As shown in Figure 8C, the thymus index was significantly lower in MM and TX mice than in MC mice (p < 0.01); Tongxie Yaofang administration produced a slight, nonsignificant recovery.

Figure 8D shows that the model induced a significant increase in the serum CRH concentration in MM mice compared with that in MC mice (p < 0.01). Treatment with Tongxie Yaofang significantly reduced CRH below MC levels (p < 0.001). As shown in Figure 8E, the serum MTL decreased in MM mice relative to MC mice (p < 0.05) and was restored to baseline following TX treatment. Figures 8F,G illustrate the changes in TBA levels. In the model group, the serum TBA level decreased, which normalized after Tongxie Yaofang administration, whereas the liver TBA level increased significantly in the MM group (p < 0.01) and returned to MC levels in the TX group.

As shown in Figure 9, HE staining of MC-group livers revealed an intact lobular architecture, clear and patent hepatic sinusoids, radially arranged polygonal hepatocytes with uniform cytoplasmic eosinophilia, distinct nuclei, and no hepatocellular swelling, degeneration, or inflammatory infiltration. In contrast, MM-group livers presented a disrupted lobular structure, hepatocellular swelling and rupture, and marked inflammatory cell infiltration. The TX group notably attenuated hepatocyte damage and reduced inflammatory infiltration, restoring near-normal lobular organization.

In the MC group spleens, the white and red pulp areas were well demarcated, with dense parenchymal organization, distinct splenic corpuscles exhibiting prominent germinal centers, and balanced white-to-red pulp proportions. MM mice displayed substantial pathology, including marked white pulp atrophy, disrupted architecture with white pulp reduced to scattered cords or small clusters, and blurred white‒red pulp boundaries. Following Tongxie Yaofang administration, the spleens from the TX group presented improved white pulp preservation, clearer demarcation between white and red pulp, and restoration of the normal splenic architecture.

The rarefaction curve illustrates how α diversity changes with sequencing depth, and its plateau indicates that additional sequencing yields few new species. As shown in Figure 10A, the curves for the MC, MM, and TX groups all levelled off, demonstrating that sequencing depth adequately captured the vast majority of the microbial taxa. The ASV/OTU UpSet plot (Figure 10B) quantified unique and shared taxa: the MC group contained 6330 ASVs (5415 unique), the MM group 4515 ASVs (3472 unique), and the TX group 5711 ASVs (4561 unique).

α-diversity indices assess species richness and evenness. The Chao1 and Observed_species indices quantify richness, while the Shannon and Simpson indices assess diversity. Figure 10C illustrates that there were no significant changes in Chao1 (p = 0.76) or Observed_species (p = 0.68) across the groups. The MM and TX groups presented nonsignificant upwards trends in the Simpson index (p = 0.44), and the Shannon index of the TX group returned to MC levels without reaching significance (p = 0.89). These results indicate that the model perturbed microbial richness and diversity and that Tongxie Yaofang partially restored these parameters.

β-diversity analysis via principal coordinate analysis (PCoA) visualizes intersample differences. Figure 10D shows that PCo1 and PCo2 explain 37.7% and 24.1% of the variance, respectively. The MC and MM groups had separate clusters, indicating that modeling substantially modified the mucosal microbiota composition. The TX group exhibited partial overlap with the MM group while leaning towards the MC cluster, suggesting that Tongxie Yaofang facilitated the regeneration of the colonic mucosal microbial population.

Figure 11A illustrates the relative abundances of the top 10 phyla and top 20 genera in the colonic mucosa prior to and following modeling and treatment. At the phylum level (Figure 11B), Firmicutes, Campylobacterota, Bacteroidota, and Desulfobacterota_I predominated. Compared with the MM group, the TX group exhibited an increase in the abundance of Campylobacterota and Bacteroidota, while demonstrating a decrease in Deferribacterota abundance. The Firmicutes and Desulfobacterota_I levels in the TX group closely resembled those in the MC group. The modeling elevated the Firmicutes-to-Bacteroidota (F/B) ratio in MM, which was diminished in the TX group, while the change was not statistically significant (p > 0.05) (Figures 11D–F).

At the genus level (Figure 11C), Blautia_A, Parabacteroides_B, and Bacteroides_H were the most abundant. As shown in Figures 11G–I, Blautia_A comprised 3.92% of the TX microbiota, which was significantly greater than that reported for MM (0.59%) and MC (0%). The Bacteroides_H in TX (7.43%) closely matched that in MM (7.22%) and far exceeded that in MC (0.53%). Parabacteroides_B rose to 4.69% in TX, paralleling the trend observed for Blautia_A, although the intergroup differences were not statistically significant (p > 0.05).

These shifts indicate that Tongxie Yaofang partially restored the dominant microbial community structure toward that of healthy controls.

Linear discriminant analysis effect size (LEfSe) was conducted to identify genera with significantly different abundances among groups (LDA score >3). Comparing MC versus MM (Figure 12A), ten bacterial genera were enriched in the MC group: Emergencia (LDA = 3.8364, p = 0.0054), Paramuribaculum (LDA = 3.6441, p = 0.0472), Cryptobacteroides (LDA = 3.6039, p = 0.0465), Muribaculum (LDA = 3.5622, p = 0.0283), Clostridium_Q (LDA = 3.3128, p = 0.0163), CAG_83 (LDA = 3.4289, p = 0.0163), CAG_95 (LDA = 3.2678, p = 0.0264), 1XD42_69 (LDA = 3.1316, p = 0.0186), UBA946 (LDA = 3.0177, p = 0.0186), and CAG_269 (LDA = 3.0150, p = 0.0343). Whereas eight bacterial genera were enriched in the MM group: Bacteroides_H (LDA = 4.5705, p = 0.0283), Mailhella (LDA = 4.2726, p = 0.0472), Phocaeicola_A (LDA = 3.8170, p = 0.0472), Paludicola (LDA = 3.6712, p = 0.0090), Ileibacterium (LDA = 3.4858, p = 0.0343), Lawsonibacter (LDA = 3.4750, p = 0.0472), Erysipelatoclostridium (LDA = 3.2184, p = 0.0132), and QAMM01 (LDA = 3.1437, p = 0.0053). When MM was compared with TX (Figure 12C), Clostridium_Q (LDA = 4.1977, p = 0.0186), Paludicola (LDA = 3.6399, p = 0.0090), and Anaerotruncus (LDA = 3.5217, p = 0.0472) remained enriched in MM, whereas Velocimicrobium (LDA = 3.5415, p = 0.0186) was the sole genus enriched in the TX group.

A random forest model ranked the top 10 discriminatory genera across MC, MM, and TX (Figures 12B,D). In the MC vs. MM comparison, Phocaeicola_A , QAMM01, Escherichia, and Limosilactobacillus exhibited importance >0.04 and higher abundance in MM. In MM vs. TX, CAG_95 and Ventrisoma scored >0.04, with CAG_95 enriched in TX and Ventrisoma enriched in MM.

Receiver operating characteristic (ROC) analysis (AUC >0.8) was used to assess diagnostic performance. In MC and MM groups (Figure 12E), the genera with an AUC >0.8 included CAG_95 (AUC = 0.92), Paludicola (AUC = 1.00), QAMM01 (AUC = 1.00), Emergencia (AUC = 1.00), Muribaculum (AUC = 0.88), UBA946 (AUC = 0.88), Limosilactobacillus (AUC = 0.84), Escherichia (AUC = 0.90), and Nanosyncoccus (AUC = 0.88). In the MM and TX groups (Figure 12F), Paludicola (AUC = 1.00), Ventrisoma (AUC = 0.82), Phocea (AUC = 0.80), CAG_95 (AUC = 0.82), and Clostridium_Q (AUC = 0.90) met the AUC criterion.

In summary, CAG_95, Paludicola, QAMM01, Emergencia, Muribaculum, and UBA946 are potential diagnostic markers for the diarrhea model, whereas Paludicola, Ventrisoma, CAG_95, and Clostridium_Q emerge as candidate therapeutic biomarkers following Tongxie Yaofang intervention.

Using PICRUSt2 and the KEGG database, we predicted functional and metabolic changes in the colonic mucosal microbiota of mice with liver hyperactivity with spleen deficiency pattern of diarrhea following Tongxie Yaofang treatment. At the secondary pathway level, 29 functional categories were identified (Figure 13A), encompassing cellular processes, environmental information processing, genetic information processing, human diseases, metabolism, and organismal systems. Notably, metabolic pathways presented the highest overall abundance.

Focusing on tertiary pathways with median abundances >431.6 (Figure 13B), the most prominent functions included amino acid metabolism, biosynthesis of other secondary metabolites, carbohydrate metabolism, and polysaccharide biosynthesis and metabolism. Further restriction to pathways with median abundances >599.2 (Figure 13C) revealed significant treatment effects. Modeling induced a marked increase in ansamycin biosynthesis (MC vs. MM, p < 0.01), which was significantly reduced by Tongxie Yaofang (p < 0.05). Conversely, the biosynthesis of vancomycin-containing antibiotics decreased in MM mice but returned to baseline in the TX group after treatment (p < 0.05).

Analysis of the microbial contributors to the altered biosynthesis pathways revealed distinct community shifts between MM and TX mice (Figure 13D). In the TX group, the pathway was predominantly associated with beneficial taxa such as Bacteroides, Lachnospiraceae, and Parabacteroides_B, whereas in the MM group, it was driven largely by Phocaeicola_A and Helicobacter_C. This shift suggests that Tongxie Yaofang promotes the enrichment of health-associated genera within key metabolic functions disrupted by liver hyperactivity with spleen deficiency pattern of diarrhea.

To investigate how characteristic genera influence host physiology and microbial metabolism, we performed Spearman correlation analysis between the top genera identified by the random forest model and various macroscopic indices (food intake, water intake, body weight, defecation frequency, and fecal moisture), as well as serum levels of MTL, CRH, and TBA. We then mapped these relationships onto a network of significantly altered KEGG pathways (Figure 14). The metabolic pathways and their numbers are shown in Supplementary Table S2.

CAG_95 was not correlated with pathway ko00790 but was strongly positively correlated with ko00400 (p < 0.01, ρ = 0.6991) and negatively correlated with ko00511 (p < 0.05, ρ = −0.5442). Ventrisoma was negatively correlated with ko00720 (p < 0.05, ρ = −0.5153) and ko00020 (p < 0.05, ρ = −0.5153) and positively correlated with ko01051 (p < 0.05, ρ = 0.5928). These results suggest that the pathways ko00400, ko00511, ko00710, ko00550, ko00270, ko00900, ko00670, ko00720, ko00020, and ko01051 are pivotal in modulating the colonic mucosal microbial composition and may underlie the therapeutic effects of Tongxie Yaofang (Figures 14A–E).

To delineate associations between characteristic genera and host indices—including MTL, CRH, serum TBA, liver TBA, and organ indices (spleen, liver, and thymus) — we constructed a Spearman correlation network (Figure 14F). In this network, each node represents one of the five principal genera (Clostridium_Q, CAG_95, Paludicola and Ventrisoma) or a host variable. Edges represent correlations: solid lines signify statistical significance (p < 0.05), dashed lines indicate nonsignificant trends, red lines indicate positive correlations, and blue lines indicate negative correlations. The node size reflects its overall connectivity and influence within the network. We found that when p < 0.05, Paludicola was positively correlated with CRH, liver TBA, and the liver index, and Ventrisoma was positively correlated with CRH. When p > 0.05, CAG_95 showed positive, nonsignificant associations with the spleen index and serum TBA and a negative trend with the liver index. Clostridium_Q was primarily positively associated with the thymus index. These patterns suggest that individual genera—and their interactions — may modulate host metabolic and immune pathways integral to the therapeutic mechanism of Tongxie Yaofang in IBS-D.

Latin pharmacognostic names: Senna alexandrina Mill. (番泻叶 fān xiè yè); Atractylodes macrocephala Koidz. (白术 bái zhú); Saposhnikovia divaricata Schischk. (防风 fáng fēng); Paeoniae Radix Alba (白芍 bái sháo); Citri Reticulatae Pericarpium (陈皮 chén pí).

General terms: liver hyperactivity with spleen deficiency pattern(脾虚湿盛 pǐ xū shī shèng).

Book titles: Danxi Xinfa (丹溪心法 dān xī xīn fǎ).