Data on the optimization of the formula of Xiaokeyinshui extract combination treating diabetes mellitus using uniform experimental design in mice

This dataset is supplementary to our accepted article in Journal of Ethnopharmacology [1]. Xiaokeyinshui (XKYS) formula, an anti-diabetic formula, was recorded in many ancient Chinese medical books. Xiaokeyinshui extract combination (XEC) originated from this ancient formula, consisting extracts of four herbal drugs, i.e., Coptidis Rhizoma, Liriopes Radix, bitter melon, and Cassiae Semen. In this study, herb extracts were prepared and mixed, producing Xiaokeyinshui extract combination (XEC). The optimized formula of XEC was also investigated via uniform experimental design. Diabetes was induced in Kunming mice, using high-sugar-high-fat diet combined with injection of streptozotocin (STZ) intraperitoneally. Different formulae of XEC were intragastrically administered to diabetic mice for 28 days. Fasting blood glucose (FBG), oral glucose tolerance test (OGTT), hemoglobin A1c (HbA1c), total cholesterol (TC), total triglyceride (TG) were measured to assess the anti-diabetic effects of each formula. Multivariate second degree polynomial model was applied in the fitting of metabolic parameters, and the extremum value of each regression model was calculated using grid algorithm. In addition, an optimized formula of XEC was subjected to validation experiment in mice model. This data could provide basis for a reasonable analysis for the optimization of the formula of XEC.

validation experiment in mice model. This data could provide basis for a reasonable analysis for the optimization of the formula of XEC.
© Value of the Data • This data presented here not only describes the design of different formulae of Xiaokeyinshui extract combination (XEC) using uniform experimental design, but also describes the effects of different formulae of XEC on diabetic mice, in which diabetes was induced with a combination of both high-sugar-high-fat diet and injection of streptozotocin. • This data presented here provides a series of regression equations using the multivariate second degree polynomial model, which provides basis on the optimization on the formula of XEC. • This data can be useful for the study design aiming to decipher the mechanisms of action of XEC in the future. • This data can be useful for researchers on traditional Chinese medicine, especially for those who focus on the design of traditional Chinese medicine formula.

Data description
The data were analyzed with IBM SPSS Statistics V22.0. Results were presented as means ± standard deviations (SD). T-test was applied in the assessment of the differences among multiple groups, with p < 0.05 as statistically significant. Table 1 provided a uniform design table U 9 (9 4 ), being the principle in the experimental design. Table 2 presented the daily dose of herb extracts for mouse in each formula of Xiaokeyinshui extract combination (XEC). Table 3 presented effects of different XEC formulae on fasting blood glucose (FBG) in mice. Table 4 presented effects of different XEC formulae on blood glucose (BG) levels in 2 h oral glucose tolerance test (OGTT) in mice. Table 5 presented effects of different XEC formulae on  1  1  2  4  7  2  2  4  8  5  3  3  6  3  3  4  4  8  7  1  5  5  1  2  8  6  6  3  6  6  7  7  5  1  4  8  8  7  5  2  9 9 9 9 9 Table 2 Daily dose of herb extracts for mouse (mg/kg/d).
hemoglobin A1c (HbA1c), total cholesterol (TC), total triglyceride (TG) in mice. In brief, Tables 3 -5 presented metabolic parameters of mice. Raw data relating to these parameters could be seen in Supplementary Materials, Tables S1-S4. Both the daily dose of each herb extract, or variables (X), and the metabolic parameters, or dependent variables (Y) were converted into normalized data before regression, eliminating the difference of units. Normalized data were obtained as followed: The data were presented as means ± SD. T-test was applied in the assessment of the differences among multiple groups. * * p < 0.01, * 0.01 ≤p < 0.05, versus DC. The data were presented as means ± SD. T-test was applied in the assessment of the differences among multiple groups. * * p < 0.01, * 0.01 ≤p < 0.05, versus DC.
The regression of normalized data was done using the software named as Data Processing System (DPS Version 7.05, Refine Information Tech. Co., China) [2] . Regression was done with the multivariate second degree polynomial model: Variables X 1 -X 4 , represent the normalized data of daily dose of four herb extracts. Y, or dependent variable, represents the normalized data of metabolic parameters. The extremum value of each regression model was calculated using grid algorithm with MATLAB 14.0.
Tables 6 -10 presented statistic parameters of different regression equations concerning change of FBG, area under curve (AUC) of 2 h OGTT, levels of HbA1C, TC and TG. R value, R a (adjusted R) value, p value and F value were calculated using DPS 7.0. For each metabolic parameter, an equation was selected as regression model for further analysis, based on a combined consideration on the R, R a , p and F value. Table 11 presented extremum values of selected equations and the corresponding optimal levels of four herb extracts. Table 12 presents predicted values of metabolic parameters using optimized daily dose of four herb extracts, and made a comparison to extremum values of selected equations.  Table 13 presented experimental results in the validation experiment. Raw data relating to validation experiment could be seen in Supplementary Materials, Tables S5-S7.
TC and TG kits were purchased from Nanjing Jiancheng Science and Technology Co. (Nanjing, China).

Animal and the establishment of diabetic model
Male Kunming mice (18-20 g) were purchased from Hubei Provincial Center for Disease Control and Prevention (No.420 0 060 0 026992). The whole process of animal experiments was conducted in Laboratory Animal Center, Tongji Medical College, HUST (SYXK (Hubei) 2016-0057). Mice were acclimatized for seven days.
After acclimatization, eight mice, being normal mice, continued to receive the standard chow diet, while others received the HSHF diet. Two weeks later, HSHF-feeding mice were kept fast for 12 h and intraperitoneally (i.p.) injected STZ (120 mg/kg, pH 4.5, in citrate buffer). Normal mice were kept fast for 12 h, and subjected to injection of citrate buffer (i.p.). One week later, the level of FBG was measured via plasma glucose test strips (Bayer, Germany), with tail-tip blood. Then, mice with FBG below 11.1 mmol/L were subjected to i.p. STZ injection (40 mg/kg) again. One week later, mice with FBG above 11.1 mmol/L were regarded as diabetic.

Experimental design based on uniform design 2.3.1. Selection of daily dose of herb extracts
In Chinese Pharmacopeia, the maximum dose of Coptidis Rhizoma (crude drug) is 5 g/d for human [4] . However, a clinical report pointed out that in clinical practice, clinicians choose a dose of 3-6 g/d in long term treatment of diabetes [5] . In this experiment, we chose 6 g/d as the maximum level of Coptidis Rhizoma (crude drug).
Cassiae Semen and Liriopes Radix have the same maximum daily dose for human, namely, 15 g/d according to Chinese Pharmacopeia [4] .
Bitter melon is not recorded in 2015 Chinese Pharmacopeia, but recorded in 2018 Hubei Provincial Quality Standards of Chinese Materia Medica, with a maximum dose at 60 g/d for human [6] .
Thus, maximum dose of each herb (crude drug, for human) was set as following: Coptidis Rhizoma, 6 g/d; Liriopes Radix and Cassiae Semen, 15 g/d; fresh bitter melon, 60 g/d. The dose of each crude drug for mouse was calculated according to the following equation: In this equation, D mouse represents the dose of crude drug for mouse; D human represents the dose of crude drug for human. Average human body weight is 70 kg. The dose conversion ratio of human to mice is 9.1 [7] .
Considering the yield of each herb extracts reported in our previous research, the maximum dose of each herb extracts for mouse was set as followed: TACR, 136 mg/kg/d; LRP, 620 mg/kg/d; BME, 139 mg/kg/d; CSE, 72 mg/kg/d [1] . For ease of calculation, the minimum dose of each herb extracts was set at 1/10 of the maximum dose with small modification. Thus, the dose range of each herb extract was finally determined as followed: TACR, 16-136 mg/kg/d; LRP, 60-620 mg/kg/d; BME, 11-139 mg/kg/d; CSE, 8-72 mg/kg/d.

Formula design based on uniform design
Twelve groups were included in this study, each group comprising of eight mice. Normal control group was abbreviated as NC, whereas diabetic control, DC. Mice in both groups were intragastrically given 0.1% water solution of sodium carboxymethyl cellulose (sodium CMC). Metformin were given at a dose of 150 mg/kg/d, with this group abbreviated as MET.
Nine XEC formulae were designed according to the principles of uniform design, abbreviated as XEC1-XEC9. Here, four variables, X 1 -X 4 , represent the daily dose of TACR, LRP, BME and CSE (mg/kg/d), respectively.
The scheme of uniform design U 9 (9 4 ) was presented in Table 1 . Combined with the U 9 (9 4 ) scheme, daily dose of herb extracts was set accordingly in Table 2 , with the ranges of daily dose of four herb extracts discussed in Section 2.3.2 .
In uniform experimental design, sampling points had a uniform distribution within the range of each factor [8] . In other words, according to the principles of uniform experimental design used in this study, U 9 (9 4 ), 9 levels distributed uniformly in the range of daily dose of herb extract. For example, the range of daily dose of TACR for mouse is 16-136 mg/kg/d. Therefore, level 1 of TACR is 16 mg/kg/d; level 2 of TACR is 31 mg/kg/d;…; while level 9 of TACR is 136 mg/kg/d. Nine levels distributed uniformly in the range. Combined with uniform experimental design table U 9 (9 4 ) and the range of each herb extract, the daily dose of each herb extract was set and presented in Table 2 . XEC1-XEC9 were prepared by mixing four herb extract in a ratio in accordance to Table 2 , and then were suspended in 0.1% sodium CMC before intragastrical administration to mice.
During the experimental process, solution was intragastrically given at 10 ml/kg/d, once a day. At the end of the study, blood was collected from the retro-orbital sinus, and mice were sacrificed thereafter. A part of blood sample was centrifugated for serum (4500 rpm, 10 min), while another part was placed in tubes containing EDTA, as whole blood.
Animal experiments were conducted under the guidance of Regulations for the Administration of Affairs Concerning Experimental Animals in Hubei Province. Experimental proce-dures were carried out with approval from the Institutional Animal Care and Use Committee, Huazhong University of Science and Technology.

Measurement of FBG and performance of OGTT
FBG was monitored every week after fasting for 12 h. Raw data relating to FBG could be seen in Supplementary Materials, Table S1. Levels of FBG in day 0 were different. Thus, regression was done with change of FBG, instead of FBG, eliminating the differences in FBG in day 0. Change of FBG was calculated as: OGTT was carried out on mice fasting for 12 h overnight, on the 24th day. Raw data relating to OGTT could be seen in Supplementary Materials, Table S2. Mice were intragastrically administered with glucose solution (2.5 g/kg). Blood glucose (BG) was monitored at three time points: 0.5, 1 and 2 h, after glucose administration. BG levels before glucose administration were regarded as 0 h. Calculation of area under curve (AUC) was conducted as followed:

Measurement of HbA1c, TC and TG
HbA1c measurement was conducted with whole blood, via Ultra2 GHb meter (Primus, USA). TC and TG in serum were analyzed using commercial kits from Nanjing Jiancheng Science and Technology Co. (Nanjing, China). Raw data relating to these parameters could be seen in Supplementary Materials, Tables S3 and S4.

Fitting and regression on the changes of FBG
As can be seen from Table 3 , mice in each group had different FBG levels in day 0. Thus, changes of FBG were applied in the regression, eliminating the influence of different FBG levels in day 0. When the change of FBG is positive, it shows a hypoglycemic effect, if negative, hyperglycemic effect.
Y represents the changes of FBG, while X 1 -X 4 represent the daily dose of four herb extracts. All variables were normalized and regression was done with DPS 7.0. Several multivariate second degree polynomial functions were obtained. R, R a , p and F value were also reported in Table 6 . Equation (1)(2)(3)(4) was selected as regression model of change of FBG for further analysis.

Fitting and regression on 2h OGTT AUC
Y AUC represents the AUC of 2 h OGTT and X 1 -X 4 represent the daily dose of four herb extracts. Regression was done as previous mentioned. R, R a , p and F value were also reported in Table 7 . The equation (2-1) was selected for further analysis.

Fitting and regression on HbA1c
Y represents the HbA1c level and X 1 -X 4 represents the daily dose of four herb extracts. Regression was done as previous mentioned. R, R a , p and F value were also reported in Table 8 . The equation  was selected for further analysis.

Fitting and regression on TC
Y represents the TC level and X 1 -X 4 represent the daily dose of four herb extracts. Regression was done as previous mentioned. R, R a , p and F value were also reported in Table 9 . The equation  was selected for further analysis.

Extremum value of selected equation and the corresponding optimized level of four herb extracts
Extremum values of selected equations and the corresponding values of four herb extracts were calculated using grid algorithm in MATLAB 14.0. Results were shown in Table 11 . For change of FBG, when the value is positive, it shows a hypoglycemic effect. In other words, a higher value means a better glycemic control. Thus, the maximum value of change of FBG was calculated. For other four parameters, the minimum values were calculated. Herein, the optimal daily dose of each herb extract was set with consideration on experimental data, regression results and calculation of extremum values. The optimized daily dose of XEC for mouse was set  Note: "-" represent the absence of the factor. The data were presented as means ± SD. T-test was applied in the assessment of the differences among multiple groups. * * p < 0.01, * 0.01 ≤p < 0.05, versus DCV. ## p < 0.01, # 0.01 ≤p < 0.05, versus METV.