During the preparation of the hydrogel patches, sensory properties are a key factor in assessing performance and user experience (Alfei et al., 2023; Klein Cerrejon et al., 2025; Li et al., 2025; Quesada-Cabrera and Parkin, 2020). The scoring criteria for the comprehensive sensory evaluation are provided in Table 1. Based on key macroscopic sensory indicators such as spreadability, uniformity, transparency, skin-following ability, and residue, a preliminary comprehensive sensory score was given to the prepared hydrogel patches (Guo et al., 2023; Wang et al., 2026). This approach provides an intuitive and comprehensive initial evaluation of their performance from a user experience perspective.

To assess spreadability is to evaluate the ease with which the patch can be applied to the skin. Good spreadability is crucial for ensuring uniform coverage of the target area, thereby enhancing its therapeutic and protective effects. Uniformity examines whether the patch’s texture is consistent, which not only affects its appearance but also the distribution of active ingredients. Breathability is essential for patches used over long periods, as it ensures proper skin respiration and reduces the risk of allergic reactions. Skin-following ability refers to the patch’s capacity to conform to the skin during movement. Good conformity ensures the patch adheres tightly to the skin, preventing it from lifting or falling off, which maintains its stability and effectiveness. Residue evaluates what is left on the skin after the patch is removed; low residue significantly improves user acceptance and satisfaction.

Evaluation of these metrics provides a multidimensional reflection of the hydrogel patch’s overall performance and user experience. The preliminary comprehensive sensory scoring allows researchers to compare and screen different batches or formulations, providing a basis for subsequent optimization and improvement. Furthermore, this evaluation method offers an important reference for product quality control and market competitiveness analysis, helping to develop hydrogel patch products that better meet user needs.

To determine the optimal amounts of poly (acrylic acid), glycerol, aluminum glycolate, sodium carboxymethyl cellulose, and tartaric acid, we prepared a blank plaster matrix (Zhao et al., 2023). The quality of the matrix was assessed using a comprehensive sensory evaluation method based on fuzzy mathematics. The evaluation described the formation of the matrix based on key characteristics such as firmness (spreadability), tackiness (uniformity), fabric penetration, film residue, and skin-following ability.

The hydrogel precursor solution and the cross-linked hydrogel were placed in separate 5 mL centrifuge tubes on an inclined plane with a 30° angle (Lu et al., 2021). After standing for 30 min, their changes in shape and displacement were recorded to assess the gel-forming state and differences in fluidity.

Hydrogel samples were applied to substrates with different surface properties, including glass, polypropylene plastic, stainless steel, ceramic, and polytetrafluoroethylene (Liu et al., 2020; Nosrati-Siahmazgi et al., 2024). The system was then inverted to observe and record whether the substrates detached under the hydrogel’s own weight, which was used to evaluate the hydrogel’s adhesion strength to materials with different friction coefficients.

A suitable amount of the hydrogel sample was uniformly applied to the dorsal side of a subject’s index finger metacarpophalangeal joint. The finger joint was then flexed and extended 50 times at a constant rate of one extension per second. After the test, the hydrogel’s adhesion status in the joint area was observed and recorded in both the upright (gel side up) and inverted (gel side down) positions.

Strip-shaped hydrogel samples, measuring 25 mm long, 10 mm wide, and 2 mm thick, were prepared. A uniaxial and cyclic tensile test was performed by holding both ends with tweezers (Li et al., 2022). The maximum tensile length at break and the residual deformation of the sample at the end of the cycle were recorded.

Two cylindrical hydrogel samples of the same size but different colors (35 mm diameter × 17.5 mm height) were prepared. Each sample was precisely cut in half along its central cross-section to obtain two pairs of semi-cylinders. The freshly cut surfaces of the different colored halves were brought into close contact and allowed to heal at room temperature for 5 min. A tensile test was then performed at the healed interface to observe the fracture location and record the healing efficiency.

The hydrogel samples were sputter-coated with 5 nm of gold-palladium (Au/Pd) using a high-vacuum ion sputter coater (Leica EM ACE600) and then imaged under high vacuum with a 5 kV accelerating voltage. A field-emission scanning electron microscope (FE-SEM, Hitachi SU8010) with 10,000x magnification was used to analyze the surface morphology and cross-sectional microstructure.

Rheological characterization was performed using a rotational rheometer. All measurements were conducted at 25 °C ± 0.1 °C.

Amplitude Sweep: The linear viscoelastic region (LVR) was determined by an oscillatory amplitude sweep (0.01%–100% strain, 1 Hz frequency). The storage modulus (G′) and loss modulus (G″) were recorded to identify the strain range where the moduli remain strain-independent.

Creep-Recovery Test: Within the LVR (at a constant stress of 200 Pa), the creep compliance (Jt) was monitored during a 300 s loading phase and a 600 s recovery phase. The elastic recovery rate (R) was calculated according to Equation 1. R % = J max − J final J max × 100 % (1)

Frequency Sweep: A viscoelastic spectrum was obtained at a constant strain of 1% across an angular frequency range of 0.1–100 rad/s. The frequency-dependent storage modulus (G′ω) and loss modulus (G″ω) were analyzed to evaluate structural relaxation dynamics.

Self-Healing Test: An intact hydrogel sample was placed on the lower plate, and the gap was adjusted to the specified thickness. A strain amplitude (γ = 1%) was applied at an angular frequency (ω = 10 rad/s) to measure the storage modulus (G′) and loss modulus (G″) for 120 s to confirm a stable gel state. A high strain (γ = 500%) was then applied for 60 s to induce shear fracture of the sample, and the plateau modulus (G) was recorded. Immediately after, the strain was returned to a low amplitude (γ = 1%), and the changes in G′ and G″ were continuously monitored over 30 min. The self-healing efficiency was calculated as the ratio of the final healed modulus (G′) to the initial modulus.

Fourier-transform infrared spectroscopy (FT-IR) was used to analyze the changes in functional groups among PAA, CMC-Na, and the cross-linking agent within the hydrogel.

First, a series of standard solutions of liquiritigenin were prepared at varying concentrations using a pH 7.4 phosphate-buffered saline (PBS) solution. The absorbance of each solution was measured at 276 nm (λ_max of liquiritigenin) using a UV-Vis spectrophotometer to construct a standard curve, which was used for subsequent quantitative analysis of drug concentration.

The prepared hydrogel patches were cut into a standard size of 2 cm × 2 cm and placed into pre-treated dialysis bags. The dialysis bags were then immersed in a conical flask containing 50 mL of release medium and transferred to a constant-temperature shaking incubator set at 37 °C with a shaking speed of 100 rpm. At predetermined time points (0.5, 1, 2, 4, 6, 8, 12, 24, 36, and 48 h), 1 mL of the release medium was sampled from the conical flask and immediately replaced with an equal volume of fresh release medium. The absorbance of the samples at each time point was measured using the UV-Vis spectrophotometer.

Based on the standard curve, the absorbance values were converted to the corresponding drug concentrations to calculate the cumulative drug release. The cumulative release rate (%) was calculated using the following Equation 2: Cumulative release rate % = Total   drug   load Cumulative   release   amount × 100 % (2)

Finally, plot the cumulative release rate as a function of time to intuitively present the drug release dynamics.

Hydrogel samples were cut into a standard size of 2 cm × 2 cm. Their initial dry weight, denoted as Wd, was precisely measured using an analytical balance. At least three parallel samples were prepared for each group. The samples were then immersed in PBS buffer solutions at pH 5.5 and pH 7.4, as well as deionized water, and placed in a constant-temperature shaking incubator at 37 °C. At predetermined time points (1, 2, 4, 8, 12, 24, 36, and 48 h), the samples were removed, excess surface water was quickly blotted with filter paper, and their wet weight, denoted as Ws, was immediately measured.

The swelling rate (%) is calculated using the following Equation 3: swelling rate  % = W s ‐ W d W d × 100 % (3)

By plotting the curve of swelling rate over time and conducting statistical analysis, evaluate the swelling behavior of the hydrogel.

Hydrogel samples were cut into a standard size of 2 cm × 2 cm. Their initial mass, denoted as W0, was precisely measured using an analytical balance. The samples were then placed in pre-weighed aluminum foil pans or Petri dishes and stored in a constant-temperature and humidity incubator set at 37 °C and a specific relative humidity. At predetermined time points (0, 1, 2, 4, 8, 12, and 24 h), the samples and pans were removed, and the total mass, denoted as Wt, was quickly measured.

The mass of each sample at each time point was calculated, and the dehydration rate (%) was determined using the following Equation 4: Dehydration Rate  % = W 0 ‐ W t W 0 × 100 % (4)

A curve plotting the dehydration rate versus time was then created, and the slope of the curve was analyzed to evaluate the hydrogel’s water retention capacity.

All molecular dynamics (MD) simulations were performed using the open-source GROMACS software package (version 2023.2). The atomic interactions within the system were described by the OPLS-AA force field. Long-range electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method, while van der Waals interactions were treated with a cut-off radius of 1.0 nm. Periodic boundary conditions were applied in all three directions.

The initial simulation box was constructed to include Polyacrylic acid (PAA), sodium hydroxymethylcellulose (CMC-Na), aluminum glycolate (alu), glycerol (Gly), ethanol (Eth), and LQ. The molecules were randomly placed within a periodic simulation box to create an amorphous cell. Prior to the MD simulation, the system underwent a two-step energy minimization process to remove any steric clashes and achieve a stable starting configuration. The steepest descent algorithm was used to minimize the system’s potential energy until the maximum force on any atom was below 1,000 kJ mol⁻¹ nm⁻¹.

Following energy minimization, the system was equilibrated under the NPT ensemble (constant particle number, pressure, and temperature). The temperature was maintained at 300.0 K using a V-rescale thermostat with a time constant of 0.1 ps. The pressure was controlled at 1.0 bar using a Parrinello-Rahman barostat with a time constant of 2.0 ps. A time step of 1 fs was used for the entire simulation, which ran for a total of 500.0 ps.

Finally, the porosity of the equilibrated system was analyzed from the simulation trajectory. The solvent-accessible surface area (SASA) was calculated using the `gmx sasa`tool, and the porosity was derived from the resulting data.

The 2D and 3D structures of liquiritigenin (LQ) were downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/), and sepsis-related target proteins were downloaded from the PDB database (https://www.rcsb.org/). Molecular docking was performed using AutoDock Vina to analyze the binding scores between liquiritigenin and the target proteins. The best-acting target was selected, and the optimal docking conformation was visualized.

3T3 Cells were seeded into 96-well plates at a density of 5 × 10³cells per well. Each well was supplemented with 100 µL of complete culture medium and cultured overnight in a cell incubator. The culture medium was removed, and each well was treated with 100 µL of hydrogel extract at different concentrations (0.1, 0.2, and 0.4 g/mL). A blank control group (medium + CCK-8, no cells),a negative control group (untreated cells) and a vehicle control group (blank hydrogel extract) were included. The 96-well plate was returned to the incubator and detected at predetermined time points, including 24, 48, and 72 h. At each time point, the treatment solution was removed, and the cells were washed once with PBS. Then, 100 µL of fresh complete culture medium and 10 µL of CCK-8 reagent were added to each well and gently mixed. The plate was incubated in the incubator away from light for 1–4 h. The absorbance value (OD value) was read at a wavelength of 450 nm using a microplate reader.

After various treatments, cell suspensions were collected and mixed with trypan blue dye solution at a 1:1 ratio. The ratio of unstained live cells to the total number of cells was then counted to assess the hydrogel’s cytotoxicity.

To investigate the hydrogel’s anti-inflammatory properties, a macrophage inflammation model was established using RAW264.7 cells induced by lipopolysaccharide (LPS) secreted by E. coli. RAW264.7 cells were seeded into culture plates, and once adherent, they were induced with a final concentration of 1 μg/mL of LPS. After different treatment durations, the cell culture supernatants were collected to quantitatively measure the expression levels of inflammatory factors (TNF-α, IL-6, TLR4) using ELISA or other immunological methods.

Log-phase E. coli and S. aureus strains were separately diluted to 10⁶–10⁷ CFU/mL with saline. A series of dilutions of the hydrogel extract were then added to culture tubes containing the bacterial suspension and incubated at 37 °C for 24 h, and then the minimum concentration required to inhibit bacterial growth was observe.

The specific experimental design and results are shown in Table 4.

Using Design-Expert 13.0 software, we performed a regression analysis of the experimental results (OD values) from Table 4. A second-order regression model was used to derive the following Equation 5: Y = 0.864 − 0.01 A − 0.03 B + 0.01 C − 0.006 AB + 0.04 AC − 0.02 BC − 0.13 A 2 − 0.17 B 2 − 0.25 C 2 (5)

An analysis of variance (ANOVA) was performed on the regression model for the response value (OD), with the results shown in Table 5. The factors aluminum Gly-Al (B) and the interactions between NP-700 (A) and glycerol (C) had a significant effect on the OD value (P < 0.05). The order of influence on the OD value was: aluminum Gly-Al (B) > glycerol (C) > PAA (A).

The regression model for the response surface had an F-value of 44.87 and a P-value of less than 0.0001, indicating that the model is highly significant (P < 0.01). Furthermore, the lack-of-fit P-value was 0.2147 (>0.05), which is not significant. This suggests that non-experimental factors have a minor effect on the OD value, and the model has good experimental stability. The regression coefficient (R2) was 0.9758, and the adjusted regression coefficient (Radj2) was 0.9541, which are very close. This demonstrates that the model possesses sufficient accuracy and generalizability. The predicted Rpre2 was 0.8619, and the difference between Rpre2 and Radj2 was less than 0.2, confirming the reliability of the predictive results. In summary, all calculated results indicate that the model is highly reliable and can be used for result analysis and condition prediction.

Based on the results of the regression model’s analysis of variance, we used Design-Expert 13.0 software and the regression equation to create two-dimensional interaction plots and three-dimensional contour plots. These plots illustrate the effects of sodium polyacrylate (A), aluminum Gly-Al (B), and glycerol (C) on the response value, OD. When one of the three factors is held constant, the interaction between the other two factors and their influence on the response value can be represented by contour lines and interaction lines. The results are shown in Figure 2.

A two-by-two interaction analysis was performed on PAA (A), aluminum glycolate (B), and glycerol (C) to create interaction plots. These plots reflect the mutual interactions between factors and help identify the optimal parameters. The steeper the surface and the denser the contour lines, the more significant the effect. The more elliptical the contour lines, the stronger the interaction between the two factors.

Based on Figure 2A, the OD value initially increases with the addition of PAA (A) before decreasing, with the rates of increase and decrease being roughly similar. In contrast, while the OD value also increases and then decreases with the addition of aluminum Gly-Al (B), the rate of decrease is much more pronounced. This indicates that changes in the aluminum Gly-Al (B) level have a more significant effect on the OD value, playing a more critical role in the interaction between the two factors. The contour lines are nearly circular, suggesting that the factor interaction is moderate. The contour plot also shows that the predicted result is highest at the central levels of PAA (A) and aluminum Gly-Al (B).

As seen in Figure 2B, the interaction between PAA (A) and glycerol (C) is significant (P < 0.05). When glycerol (C) is at a low or high level, increasing PAA (A) leads to two completely opposite trends: a slow increase followed by a rapid decrease, and a rapid increase followed by a slow decrease, respectively. Similarly, when PAA (A) is at different levels, increasing glycerol (C) also produces entirely opposite results. These trends, combined with the very steep response surface and elliptical contour lines, confirm a significant interaction between the factors, which has a major impact on the results. The contour plot shows that a high predicted value (above 0.8) can be achieved when PAA (A) is between 3.8 and 6.0 g and glycerol (C) is between 31 and 50 g.

Figure 2C shows that the OD value slowly increases and then rapidly decreases as aluminum Gly-Al (B) is increased, with a clear difference in the rates of change. However, as glycerol (C) increases, the OD value also rises before falling, but the rates of change are consistent. This comparison indicates that the result is more significantly affected by changes in aluminum Gly-Al (B), while glycerol (C) plays a more dominant role in the interaction process. This finding is consistent with the influence order observed in the single-factor analysis. The contour plot reveals that high predicted values are obtained when aluminum Gly-Al (B) is between 0.48 and 0.68 g and glycerol (C) is between 30 and 50 g.

Figure 3A’s single-factor “awakening plots” (also known as perturbation plots) visually show the effect of factors A, B, and C on the OD value. All three response curves exhibit a “rise then fall” pattern, which is a parabolic shape. This confirms the highly significant quadratic terms (A2, B2, and C2) in the regression equation (P < 0.0001). This indicates a significant quadratic effect of each factor on the OD value, meaning the relationship between the factor levels and the OD value is non-linear. An optimal level exists for each factor, beyond which the OD value decreases as the factor level increases.

The interaction plot in Figure 3B further quantifies the strength of the factor interactions. The interaction curves for A and C show significant separation, indicating a clear difference in the trend of OD value changes across different level combinations. This further confirms the significance of their interaction. In contrast, the interaction curves for A and B, and B and C, nearly overlap with minimal separation. This shows that the interaction between these two pairs of factors has no significant effect on the OD value, which is consistent with the conclusion drawn from the contour plots.

The perturbation plot in Figure 3C uses the center level as a reference point to show how the OD value changes as factor levels deviate from the center. As A, B, and C deviate toward the star points, the OD value consistently decreases. This visually confirms that the central region is the optimal response range for the OD value, which aligns with the characteristic of the response surface plot where the highest OD value is found in the central region.

Figure 3D shows the scatter plot of predicted versus actual values. The data points are tightly clustered around the diagonal line, indicating a high degree of consistency between the model’s predicted OD values and the actual experimental values. This confirms the accuracy of the regression model in predicting the OD value.

In the 3D residual plot in Figure 3E, the residual points are randomly distributed near the zero-value plane, with no signs of clustering or a regular pattern. According to the principle that “randomly distributed residuals indicate no systematic error and a sufficient fit,” this shows that the residuals of the regression model are random errors. The model provides a sufficiently reliable fit for the experimental data without introducing any systematic bias.

In summary, the second-order regression model can accurately fit the relationship between PAA, aluminum glycolate, glycerol, and the OD value. Only the interaction between NP-700 and glycerol had a significant effect on the OD value. All factors showed a significant quadratic effect on the OD value, with clear optimal level ranges. The model’s predictions for the OD value are highly accurate and reliable, providing a strong theoretical basis for future process optimization.

To maximize the response value (OD), an effective regression model was used to optimize the conditions and find the optimal solution. The optimization objective function and mathematical model are as follows: max{Y(OD value), A∈[2.56, 7.44], B∈[0.317, 0.900], C∈[10, 70]}.

By solving the mathematical model, we found the optimal conditions for each factor: PAA (A) at 4.932 g, aluminum glycolate (B) at 0.582 g, and glycerol (C) at 40.658 g. Considering practical experimental conditions and feasibility, these parameters were set to PAA (A): 4.90 g, aluminum glycolate (B): 0.58 g, and glycerol (C): 41 g.

A validation test was conducted with these parameters in three repeated experiments. The average OD value obtained was 0.939 ± 0.001, indicating a good result. The specific experimental data are shown in Table 6. This study validates the accuracy and reliability of the model, providing optimal process parameters that can be applied in practice.

The NPT molecular dynamics simulation of the hydrogel system (“LQ”) was conducted for 500 ps to evaluate its thermodynamic stability and drug transport properties. As shown in Figure 6, the system density stabilized at approximately 1.2 g/cm³, and the total energy remained conserved with controlled temperature fluctuations, confirming that the system reached a stable equilibrium suitable for analysis. Crucially, the simulation provided molecular-level insights into the drug release mechanism observed in vitro. The Mean Square Displacement (MSD) analysis yielded a diffusion coefficient (D) for liquiritigenin (LQ) of 1.202 times 10^–3 Ų/ps. This relatively low diffusion coefficient indicates that the dense, cross-linked PAA/CMC-Na network significantly restricts the mobility of the drug molecules. This physical parameter correlates directly with the sustained release profile shown in Figure 5G, providing a theoretical basis for the hydrogel’s ability to maintain therapeutic drug concentrations over 24 h without an immediate, uncontrolled burst release. Furthermore, the Radial Distribution Function (RDF) analysis revealed distinct peaks in the interaction probability between Aluminum ions (Al) and Carboxyl groups (COO⁻) within the 0–10.5 Å range. These strong specific interactions confirm the formation of stable coordination bonds acting as physical cross-links. Biologically, this high cross-linking density is essential for maintaining the structural integrity of the hydrogel patch in physiological environments, ensuring that it remains adhered to the wound bed and resists rapid erosion by wound exudates, thereby fulfilling the requirements for a long-lasting wound dressing.

Potential therapeutic targets for sepsis, including Toll-like receptor 4 (TLR4), tumor necrosis factor-alpha (TNF-α), plasminogen activator inhibitor-1 (PAI-1), MAPK, and high-mobility group box 1 (HMGB1), were identified from the Protein Data Bank (PDB). Molecular docking of LQ to these target (Figure 7; Table 7) suggests a multi-mechanistic anti-sepsis effect, likely involving TLR4, TNF-α, interleukin-1 β (IL-1β), HMGB1, PAI-1, and p38 MAPK. The highest binding affinity was observed with TLR4 (Score: −7.085 kcal/mol), indicating LQ may act as a potent TLR4 antagonist to competitively inhibit lipopolysaccharide (LPS) binding, thereby blocking the inflammatory signal initiation. Furthermore, LQ displayed favorable affinities for key pro-inflammatory mediators, including TNF-α (−6.434 kcal/mol), IL-1β (−5.586 kcal/mol), and the late-release DAMP, HMGB1 (−5.832 kcal/mol). This suggests LQ can neutralize these mediators or interfere with their receptor interactions, thus suppressing the systemic inflammatory cascade. As a flavonoid, LQ is also proposed to modulate downstream signaling pathways, such as p38 mitogen-activated protein kinase (p38 MAPK), to inhibit the activation of inflammatory transcription factors and reduce mediator synthesis. Finally, LQ’s potential binding to PAI-1 suggests a beneficial role in mitigating sepsis-induced disseminated intravascular coagulation (DIC). Collectively, LQ shows promise as a comprehensive sepsis therapeutic via the synergistic mechanisms of inhibiting inflammation initiation, blocking signal transduction, and neutralizing inflammatory mediators.