Analysis Method for the Correlation between the Apparent Crystal Forms of Mixed Crystal Polymorphic Drugs and Impurity Formation

Abstract

The present invention relates to a method for analyzing the correlation between the apparent crystal form and impurity formation of a mixed crystal type of polymorphic drug; first, a method for obtaining the "apparent crystal form" of a drug through near-infrared spectroscopy (NIR) is provided, so as to be able to quickly and massively obtain the data basis of numerous marketed drugs; the canonical correlation analysis method is used to explore the canonical correlation between the macroscopic apparent crystal form and the impurity spectrum, excavate the multi-dimensional complex correlation between the population apparent crystal form and the impurities of individual drugs in the population, and then use the individual correlation transfer information in the population to establish the transfer relationship between the individual crystal form and the individual impurities, and find the correlation between the impurity category and the crystal form category. Obtaining the key point information of drug quality and process control related to the crystal form can not only realize the data collection of the quality control of this type of drug and the stability, safety and effectiveness in drug research and development, but also help to trace and study the causes and mechanisms of adverse reactions.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and particularly relates to a method for analyzing the correlation between the apparent crystal form and impurities of a mixed crystal type of polymorphic drugs. Background Art

[0002] In the process of modern drug research and development and production, drug quality control has always been a crucial link. With the progress of technology and the development of data analysis technology, drug quality research oriented by big data has gradually become a research hotspot and an important means for drug research and development, drug safety and efficacy evaluation. However, the research on big data of drug quality belongs to a frontier interdisciplinary field and is still in its infancy. There is still a lack of effective big data analysis methods to solve specific problems, making big data unable to play the expected role in the actual work of drug research.

[0003] Many drugs exhibit polymorphism. Polymorphic drugs not only differ in solid physical and chemical properties but also have a significant impact on the stability, bioavailability, and efficacy of drugs. Polymorphic drugs refer to drugs with the same chemical composition that can exist in different crystal structures in the drug. Different crystal forms of drugs not only affect the solubility, stability, and bioavailability of drugs but may also lead to the generation of different impurities during the production and storage of drugs. The presence of impurities not only reduces the purity of drugs but may also cause adverse reactions, seriously affecting the safety and efficacy of drugs.

[0004] There must be a certain correlation between the crystal form and impurities of polymorphic drugs, which has a profound impact on drug safety, efficacy, and quality controllability. It has become a research direction that has received much attention and has not only important theoretical value but also broad application prospects. Through in-depth research on the correlation between polymorphic drugs and impurities, the purity and stability of drugs can be effectively improved, the adverse reactions of drugs can be reduced, and the quality and safety of drugs can be enhanced. In addition, it will also provide strong support for the patent protection of drug crystal forms, helping pharmaceutical companies gain advantages in the fierce market competition.

[0005] Polymorphic drugs include single crystals, co-crystals, mixed crystals, etc. Currently, commonly used crystal form characterization methods include scanning electron microscopy, infrared spectroscopy, Raman spectroscopy, X-ray single crystal diffraction, X-ray powder diffraction, DSC, etc. In the case of having a single crystal form standard sample, the XRD method and molecular spectroscopy method can determine the type and proportion of crystal forms.

[105] For most drugs with polymorphism, it is often very difficult to obtain crystal form standard samples. However, these methods can only characterize single crystals or drugs with known crystal compositions. Usually, a single specific crystal needs to be cultivated to carry out relevant research, and then the correlation between different crystal forms (crystals) and impurities is analyzed and evaluated one by one.

[0006] For most drugs, they usually exist in the form of mixed crystals, and it is usually difficult to obtain single crystals. Although microarea characterization can be carried out by means of chemical imaging at present, since single specific crystals cannot be obtained, the correlation analysis and evaluation between mixed crystals and impurities cannot be carried out. It is difficult to analyze which crystal is more stable and which crystal is related to which specific impurity in many mixed crystal drugs, which poses risks to the safety, effectiveness and quality supervision of such mixed crystal drugs.

[0007] At present, there is still a lack of effective technical means and methods for the crystal form evaluation of mixed crystal polymorphic drugs, resulting in serious deficiencies in data related to the stability, safety and effectiveness of the crystal form of such drugs in drug quality control and drug research and development, and many problems such as the adverse reaction mechanism and cause cannot be traced well. Summary of the Invention

[0008] To solve the above technical problems, the present invention provides an analysis method for the correlation between the apparent crystal form and impurity formation of mixed crystal polymorphic drugs.

[0009] The technical solution adopted by the present invention is: an analysis method for the correlation between the apparent crystal form and impurity formation of mixed crystal polymorphic drugs, and the specific steps are as follows:

[0010] S1: Perform apparent crystal form detection on the target drug through an apparent crystal form characterization model to obtain big data of apparent crystal forms; screen the common impurity categories of the target drug to obtain big data of the impurity detection results of the target drug;

[0011] S2: Correlate the apparent crystal form information with the impurity information, and use the canonical correlation analysis method to analyze the correlation coefficients between the apparent crystal form and the impurities, including the canonical correlation coefficients between canonical variable pairs and the correlation coefficients between the original variables and the canonical variables;

[0012] S3: Determine the transfer relationship according to the correlation coefficients between the impurities, canonical variable pairs and crystal forms, and screen the correlation degrees between specific categories of apparent crystal forms and specific categories of impurities.

[0013] Preferably, in step S1, the apparent crystal form characterization model is established as follows:

[0014] S1.1: Screen typical samples, detect each typical sample by MDSC to obtain the apparent crystal form model of each typical sample; establish the spectral model of each typical sample by near-infrared spectroscopy; establish the connection between the apparent crystal form model and the spectral model corresponding to the same typical sample;

[0015] S1.2: Perform near-infrared spectral characterization on the sample to be detected, and combine the corresponding relationship between the spectral model and the apparent crystal form model to obtain big data information of the apparent crystal form.

[0016] Preferably, the canonical variable pairs include impurity canonical variables and apparent crystal form canonical variables, and they are in a paired relationship.

[0017] Preferably, the specific analysis steps of step S2 include

[0018] S2.1: Based on the contents of various impurities and the contents of various apparent crystal forms as data, extract multiple pairs of canonical variables, and calculate the correlation coefficients between each pair of canonical variables;

[0019] S2.2: Calculate the conversion coefficients of various impurities and impurity canonical variables, and various apparent crystal forms and apparent crystal form canonical variables through the canonical correlation method; generate the relational formula of each canonical variable;

[0020] S2.3: Test the relational formula of the generated pair of canonical variables, and screen out the pair of canonical variables with a p-value less than 0.05;

[0021] S2.4: Respectively construct the correlation coefficient matrices of the multiple pairs of canonical variables obtained by screening with impurities and apparent crystal forms, and obtain the correlation coefficients between impurities, each canonical variable and apparent crystal forms.

[0022] Preferably, the canonical correlation coefficients are tested by the likelihood ratio method to screen out the pairs of canonical variables.

[0023] Preferably, in step S2.4, calculate the correlation coefficient matrix of impurity-impurity canonical variables, the correlation coefficient matrix of apparent crystal form-apparent crystal form canonical variables, the correlation coefficient matrix of apparent crystal form-impurity canonical variables, and the correlation coefficient matrix of impurity-apparent crystal form canonical variables respectively.

[0024] Preferably, step S3 includes

[0025] S3.1: Screen out the correlation coefficients not less than 0.3 as alternative transfer relationships;

[0026] S3.2: It is determined that the apparent crystal forms and impurities with the correlation coefficients of the same group of typical impurity variables being all positively correlated or all negatively correlated are correlated.

[0027] Preferably, in step S3.1, screen out the correlation coefficients not less than 0.3 in the correlation coefficient matrix of apparent crystal form-impurity canonical variables and the correlation coefficient matrix of impurity-apparent crystal form canonical variables, and screen out the corresponding impurities or apparent crystal forms;

[0028] In step S3.2, based on the correlation coefficients in the correlation coefficient matrix of impurity-impurity canonical variables and the correlation coefficient matrix of apparent crystal form-apparent crystal form canonical variables, compare whether the impurities and apparent crystal forms for the same pair of canonical variables are all positively correlated or all negatively correlated.

[0029] Preferably, the correlation coefficients of the correlated impurities and apparent crystal forms are equal to the canonical correlation coefficients between the corresponding pairs of canonical variables.

[0030] Preferably, the screened apparent crystal form with no or low correlation with impurities is the most stable crystal form.

[0031] The advantages and positive effects of the present invention are as follows: "apparent crystal form" is defined from a macroscopic perspective, and effective apparent crystal form response categories and crystal characterization methods are jointly established by applying a variety of crystal form analysis methods, and then a big data analysis method based on apparent crystal form and impurity spectrum is established;

[0032] On the one hand, the multi-dimensional correlation correspondence relationship between crystal form and impurities is obtained, and then the relatively stable crystal form information in the mixed crystal type of polymorphic drugs is analyzed; on the other hand, the key point information of drug quality and process control related to crystal form is obtained, which can not only realize the data collection of the quality control of such drugs and the stability, safety and effectiveness in drug research and development, but also help to trace and study the causes and mechanisms of adverse reactions; the relatively most stable crystal form in the mixed crystal can be screened through the correlation between apparent crystal form and impurity spectrum. Description of the Drawings

[0033] Figure 1 Analysis method flow of the relationship between apparent crystal form and impurities;

[0034] Figure 2 Infrared spectra of cefathiamidine for injection of GBZ, FJFA and SDLX;

[0035] Figure 3 XRD spectra and characteristic peaks of cefathiamidine for injection; (a): GBZ; (b): FJFA; (c): SDLX; (d): GBTX;

[0036] Figure 4 TGA and MDSC spectra of cefathiamidine for injection of 4 types of apparent crystal form response categories; (a): GBZ; (b): FJFA; (c): SDLX; (d): GBTX;

[0037] Figure 5 Correlation diagram of the first canonical variable pair of crystal-impurity correlation;

[0038] Figure 6 Correlation diagram of the second canonical variable pair of crystal-impurity correlation;

[0039] Figure 7 Transfer diagram of the correlation relationship between crystal form-impurity variables and canonical variables. Detailed Embodiments

[0040] The embodiments of the present invention will be described below with reference to the drawings.

[0041] The present invention relates to a method for analyzing the correlation between the apparent crystal form and impurity formation of a mixed crystal polymorphic drug. First, a method for obtaining the "apparent crystal form" of a drug by near-infrared spectroscopy (NIR) is provided, so that the basic data of many distributed drugs can be obtained quickly and in large quantities. Using the thinking of big data analysis, firstly, the typical correlation between the macroscopic apparent crystal form and the impurity spectrum is used to explore the multi-dimensional complex correlation between the apparent crystal form of the group and the impurities of individual drugs in the group, and then the transmission relationship between the individual crystal form and the individual impurity is established by using the individual correlation delivery information in the group, and the correlation between the impurity category and the crystal form category is found. Furthermore, by analyzing the correlation between the apparent crystal form category and the impurity category, the relatively most stable crystal form can be reversely screened, breaking through the technical limitations of traditional methods.

[0042] In order to test a large number of drug samples and obtain crystal form data, the present invention proposes the concept of "apparent crystal form" of polymorphic drugs, which solves the problem that the crystal form characteristics of mixed-crystal polymorphic drugs are difficult to characterize. Effective information acquisition is the key to big data analysis. By establishing a NIR apparent crystal form classification and characterization method based on typical MDSC results, and promoting the acquisition of crystal form-related big data of different batches of drugs from different manufacturers, it is helpful to conduct typical correlation analysis of polymorphic characteristics and impurity spectra, and obtain the corresponding relationship between crystal form and impurities.

[0043] The apparent crystal form not only reflects the control level of the drug production process, but also has a close relationship with the stability of different raw material processes. According to the results of the apparent crystal form characterization, the drug impurity spectrum is used as a set of variables, and the apparent crystal form is used as another set of variables. The canonical correlation analysis method is used to explore the relationship between the two sets of variables, analyze the crystal form impurities, and explore the correlation between the apparent crystal form and impurities. Analyze the common impurity types of drug samples, associate the mixed crystal characteristic information of the group drugs (products of representative manufacturers) characterized by "apparent crystal form" with the specific impurity information of individual drugs (each batch of products) in the group drugs, and use the canonical correlation analysis method to explore the multi-dimensional complex correlation between the apparent crystal form of the group and the impurities of individual drugs in the group. Analysis methods such as Figure 1 The specific analysis method steps are as follows.

[0044] S1: Use near-infrared spectroscopy to establish an apparent crystal form characterization model, characterize the polymorphic phenomenon in drugs from multiple manufacturers, and obtain apparent crystal form big data.

[0045] First, typical samples are screened, and each typical sample is tested by modulated differential scanning calorimetry (MDSC) to obtain the apparent crystal form model of each typical sample; the spectrum model of each typical sample is established by near-infrared spectroscopy; the apparent crystal form model and the spectrum model of the same typical sample are linked; then the samples to be tested are characterized by near-infrared spectroscopy, and the apparent crystal form big data information is obtained by near-infrared spectroscopy based on the correspondence between the spectrum model and the apparent crystal form model;

[0046] In big data analysis, the acquisition of big data information is crucial. Although the MDSC method can effectively characterize the characteristics of mixed crystals, the experimental conditions are relatively harsh. Using only experimental methods is time-consuming and cumbersome, and it is impossible to collect big data on crystal forms. Near-infrared spectroscopy has the characteristics of being fast and non-destructive, and has good popularization, but it cannot directly represent the crystal form composition of mixed crystals. Establishing an apparent crystal form model of near-infrared spectroscopy-MDSC can determine the crystal form category through near-infrared spectroscopy, so as to judge which category of apparent crystal form response category the drug sample belongs to based on the near-infrared spectroscopy results. Such a detection method has accurate results, is suitable for the collection of big data related to crystal forms, and the results can also be used for further big data analysis.

[0047] From a macroscopic perspective, "apparent crystal form (ACF)" is defined as the single crystal composition and its relative relationship in the mixed crystals of polymorphic drugs and other population mixed crystal characteristics. Using the characterization results of the MDSC for the multi-crystal composition in the mixed crystals of products from typical manufacturers, a classification model of near-infrared spectroscopy is established, and classification based on macroscopic crystal form characteristics is carried out at the molecular level. The model is verified using the investigated raw material drug sources, and a near-infrared apparent crystal form characterization model is established. This model can be generalized to predict the apparent crystal forms of all manufacturers, and can achieve the classification and characterization of crystal information in mixed crystals in the absence of a single crystal form standard product, and obtain basic big data information related to crystal forms.

[0048] S2: Screen common impurity categories of drugs.

[0049] S3: Correlate the apparent crystal form information with the impurity information, and use the canonical correlation analysis method to analyze the correlation coefficient between the apparent crystal form and the impurities. Based on the content of various impurities and the content of various apparent crystal forms, multiple pairs of canonical variables are extracted, and the correlation coefficient between each pair of canonical variables is calculated; through the canonical correlation method, the conversion coefficients between various impurities and canonical impurity variables, and between various apparent crystal forms and canonical apparent crystal form variables are calculated; the relationship formula for each canonical variable is generated.

[0050] Canonical correlation analysis is carried out on the apparent crystal form variable group and the impurity spectrum variable group. Taking the apparent crystal form factor as the independent variable X and the impurity factor as the dependent variable Y, according to the correlation relationship between the two groups of variables, canonical variables are obtained to replace the apparent crystal form and the impurities, and then the relationship between the apparent crystal form and the impurities is projected onto the relationship of the canonical variable pairs, and the conversion coefficients between the original variables and the canonical variables are obtained. The calculation formulas for the canonical apparent crystal form variable U and the canonical impurity variable V are respectively:

[0051] (Formula 1);

[0052] (Formula 2);

[0053] where \(i = 1, 2, \ldots, M\), \(M\) is the number of common factors, and \(j = 1, 2, \ldots, N\), \(N\) is the number of observed samples.

[0054] Test the generated canonical variable relationship formula. Through the correlation analysis and hypothesis testing between canonical variables, study the correlation between apparent crystal forms and impurities. Assume \(H_0\) that the canonical correlation coefficients of the current and subsequent canonical variable pairs are all 0. Use the likelihood ratio method to test the canonical correlation coefficients, and the obtained likelihood ratio statistic approximately follows the F-distribution and \(\chi\) 2 distribution. Screen the canonical variable pairs with p-value less than 0.05.

[0055] Construct the correlation coefficient matrices between the multiple canonical variable pairs obtained by screening and impurities and crystal forms respectively, and obtain the correlation coefficients between impurities, canonical variable pairs, and apparent crystal forms. Analyze the canonical correlation structure, explore the transfer relationship of the original variables - canonical variable pairs - original variables, and further evaluate the statistical significance of \(r_1\) and \(r_2\). The canonical correlation structure is the pairwise correlation coefficient matrix between each original variable and the canonical variable. Generally, it is considered that there is a substantial correlation relationship when the absolute value of the correlation coefficient is greater than 0.30, a significant correlation relationship when it is greater than 0.50, and a highly relevant relationship when it is greater than 0.80.

[0056] S4: Determine the transfer relationship according to the correlation coefficients between impurities, canonical variable pairs, and apparent crystal forms, and screen the correlation degree between specific categories of impurities and specific categories of apparent crystal forms.

[0057] The correlation between the original variables and the canonical variables is mainly transmitted through \(r_1\). Draw the transfer diagram of the correlation between each variable obtained by canonical correlation analysis and the canonical variables, including the correlation between the impurity canonical variable and the apparent crystal form canonical variable, the correspondence between the impurity and the impurity canonical variable, and the correspondence between the apparent crystal form and the apparent crystal form canonical variable. Among them, the correlation transfer route with consistent positive and negative correlations is the process control route that should be focused on. Use the transfer relationship of the significant correlation relationship route to analyze the correspondence between the apparent crystal form and the impurity.

[0058] In the model embodiment of the present invention, the relatively most stable crystal form can also be screened by using the transfer relationship between the apparent crystal form and the impurity. The apparent crystal that has no correlation with all impurities is regarded as the relatively most stable crystal form within the scope of the existing apparent crystal form model.

[0059] The present invention uses a variety of crystal form analysis methods to jointly establish an effective apparent crystal form response category and crystal characterization method; furthermore, a big data analysis method based on apparent crystal form and impurity spectrum is established. On the one hand, the multi-dimensional correlation correspondence between crystal form and impurities is obtained, and then the relatively stable crystal form information in mixed crystal type polymorphic drugs is analyzed; on the other hand, the key point information of drug quality and process control related to crystal form is obtained, which can not only realize the collection of data on the quality control of such drugs and the stability, safety and effectiveness in drug research and development, but also help to trace and study the causes and mechanisms of adverse reactions.

[0060] Taking multiple distribution products of cefathiamidine for injection prepared by manufacturers GBZ, FJFA, SDLX and GBTX as examples of the original sources, the method of the present invention is described.

[0061] Example 1: Apparent crystal form characterization method

[0062] Cefathiamidine for injection prepared by manufacturers GBZ, FJFA and SDLX was respectively subjected to infrared spectroscopy detection, and the infrared characteristic results are as Figure 2 shown, where 820 - 880 cm -1 is the crystal form characteristic spectral region. Standard samples of cefathiamidine for injection of models GBZ, FJFA, SDLX and GBTX were detected by XRD, and the XRD results are as Figure 3 shown. Since it is a mixed crystal, the XRD spectral characteristics are homogenized, and the XRD characteristic peak information of different crystals cannot be obtained. Standard samples of cefathiamidine for injection of models GBZ, FJFA, SDLX and GBTX were subjected to thermogravimetric analysis (TGA) and modulated differential scanning calorimetry (MDSC), and the results are as Figure 4 shown. It can be seen that MDSC can distinguish 4 different types of crystal form samples, and the collection of crystal form big data can be realized by MDSC detection of batch samples.

[0063] The characterization of the multi-crystal composition in the typical cefathiamidine for injection mixed crystal by the MDSC method, the results are shown in Table 1 and Figure 4 shown. Then, near-infrared spectroscopy detection was carried out on each typical sample, and a near-infrared apparent crystal form characterization model was established after one-by-one comparison. The apparent crystal form of cefathiamidine for injection from different manufacturers was characterized by the near-infrared spectroscopy detection method, and the results are shown in Table 2 and Figure 2 shown; then, the raw material supply situation of each enterprise was investigated, and the results shown in Table 2 indicate that the samples corresponding to the same raw material production enterprise belong to the same apparent crystal form response category. It shows that the crystal form information of drugs from different manufacturers can be accurately obtained through this apparent crystal form characterization method.

[0064] Table 1 MDSC crystal characterization results of four types of apparent crystal form response categories (s - strong peak, w - weak peak);

[0065]

[0066] Table 2 Sources of raw materials of cefathiamidine for injection, near-infrared characterization results and apparent crystal form response categories

[0067]

[0068] Analyze the impurity content information in more than a hundred batches of drugs, analyze the apparent crystal form information through MDSC, establish the data basis as shown in Table 3, and analyze the relationship between the apparent crystal form-impurity spectrum.

[0069] Table 3 Data on impurity content and apparent crystal form information in 117 batches of cefathiamidine for injection

[0070]

[0071]

[0072]

[0073] Comparative Example 1: Explore the correlation relationship of the impurity spectrum and the correlation relationship of the apparent crystal form in the product through simple correlation analysis

[0074] The results of the simple correlation analysis of impurities are shown in Table 4. The correlation between the variables within the impurity spectrum group is weak. Among them, there is a weak correlation between impurity A and impurity B, and between impurity E and the largest single other impurity. The results of the simple correlation analysis of the apparent crystal form are shown in Table 5. There is a certain correlation between the variables within the apparent crystal form group. Among them, crystal II and crystal Ia show a negative correlation with a correlation coefficient of -0.8295, and crystal III and crystal I show a negative correlation with a correlation coefficient of -0.7846. Crystal II and crystal III are defined as the high melting point crystal group, and crystal Ia and crystal I are defined as the low melting point crystal group. The correlation analysis results indicate that there may be crystal transformation between the two crystal groups. And there is a medium positive correlation between the apparent crystal forms within the two crystal groups. The correlation coefficient between crystal Ia and crystal I is 0.3270, and the correlation coefficient between crystal II and crystal III is 0.5067, indicating that the crystals within the group may be in a symbiotic relationship.

[0075] However, obviously the information obtained from the above simple correlation analysis is limited and cannot give information on crystal form stability.

[0076] Table 4 Results of simple correlation analysis of the impurity spectrum group

[0077]

[0078] Table 5 Autocorrelation results of the apparent crystal form group

[0079]

[0080] Example 2 Big Data Analysis Method for Apparent Crystal Form-Impurity Profile of Cefathiamidine for Injection

[0081] 2.1 Big Data Sorting and Acquisition of Cefathiamidine for Injection

[0082] The impurity profile variable group of cefathiamidine for injection contains 5 variables, namely impurity A, impurity B, impurity D, impurity E and other maximum single impurities; the apparent crystal form variable group contains 4 variables as shown in Table 1, specifically including crystal Ia, crystal I, crystal II and crystal III characterized by MDSC. The total number of observed samples (N) is 117; the statistical data of the original variables are shown in Table 6, where the mean of the apparent crystal form represents the calculation result of semi-quantitative assignment of the MDSC peak ratio and after standardization processing.

[0083] Table 6 Statistics of Original Variables Related to Apparent Crystal Form-Impurity of Cefathiamidine

[0084]

[0085] 2.2 Establishment of Big Data Analysis Method

[0086] The results of the canonical correlation coefficient, cumulative contribution rate and conversion coefficient of cefathiamidine for injection are shown in Table 7. A total of 4 pairs of common factors are extracted. The canonical conversion coefficient of crystal form factor X is A, and the canonical conversion coefficient of impurity factor Y is B. Among them, the correlation coefficient r of the first canonical variable pair is 0.5228, and the canonical determination coefficient r 2 is 0.2733, and the contribution rate reaches 58.82%. The correlation coefficient of the second canonical variable pair is 0.3300, and the contribution rate is 22.13%; the correlation coefficient of the third canonical variable pair is 0.2818, and the contribution rate is 9.9%; the correlation coefficient of the fourth canonical variable pair is 0.1150, and the contribution rate is 7.84%. The cumulative contribution rate of the four groups of canonical variable pairs is 98.68%, that is, the variation that can be explained accounts for 98.68% of the total variation. Among them, the first and second canonical correlation coefficients play a major role, explaining 80.95% of the variation, and the third and fourth canonical correlation coefficients play a relatively small role.

[0087] Table 7 Results of Canonical Correlation Analysis Related to Crystal-Impurity of Cefathiamidine

[0088]

[0089] The relationship of the second pair of canonical variables is shown in Formulas 4-6.

[0090] (Formula 3);

[0091] (Formula 4);

[0092] (Formula 5);

[0093] (Formula 6).

[0094] The correlation diagrams of the first pair of canonical variates and the second pair of canonical variates are as Figure 5 Figure 6 shown, indicating that the correlation between the two pairs of canonical correlation variables is obvious.

[0095] 2.3 Hypothesis Testing in Big Data Analysis

[0096] Hypothesis testing is carried out on the canonical correlation analysis. Assuming that the current and subsequent canonical correlation coefficients are 0, this paper uses the likelihood ratio method to test the canonical correlation coefficients. The obtained likelihood ratio statistic approximately follows the F distribution and χ 2 distribution. The hypothesis testing results of cefathiamidine for injection are shown in Table 8. The first column tests whether r1~r4 are 0. The F test shows F = 3.116 and p = 0.00001 < 0.01. The test has a highly significant difference from the hypothesis, so H0 does not hold, and at least one of r1~r4 is not 0. The second column tests whether r2~r4 are 0. The F test shows F = 2.001 and p = 0.0241 < 0.05. The test has a significant difference from the hypothesis, so H0 does not hold, and at least one of r2~r4 is not 0. The p-values of the F tests in the third and fourth columns are both greater than 0.05, so H0 holds. The χ2 test results are consistent with the F test. Therefore, it can be confirmed that r3 and r4 have no statistical significance, while r1 and r2 have certain statistical significance.

[0097] Table 8 Likelihood Ratio Test Statistics for the Canonical Correlation between the Crystal and Impurities of Cefathiamidine for Injection

[0098]

[0099] 2.4 Multidimensional Correlation Results of Big Data Analysis

[0100] The canonical correlation structure of cefathiamidine for injection is shown in Table 9.

[0101] Table 9 Canonical Correlation Structure between the Crystal Form and Impurities of Cefathiamidine

[0102]

[0103] According to the data of the correlation coefficient matrix of the apparent crystal form - impurity canonical variates and the correlation coefficient matrix of the impurity - apparent crystal form canonical variates, it can be seen that Crystal Ia, Crystal II, and Crystal III are substantially correlated with the impurity canonical variate 1; Impurity B and Impurity E are substantially correlated with the apparent crystal form canonical variate 1; and there are no variables with substantial correlation in the canonical variate 2. As Figure 7As shown, according to the data of the impurity-impurity canonical variate correlation coefficient matrix and the apparent crystal form-apparent crystal form canonical variate correlation coefficient matrix, it can be seen that crystal II and crystal canonical variate 1 have a positive correlation coefficient, and impurity B and impurity canonical variate 1 also have a positive correlation coefficient. The correlation directions of the two are the same. Therefore, crystal II and impurity B are associated through the first pair of canonical variates and have a significant correlation; crystal Ia, crystal III and apparent crystal form canonical variate 1 have a negative correlation coefficient, and impurity E and impurity canonical variate 1 also have a negative correlation coefficient. The correlation directions of the two are the same. Therefore, crystal Ia, crystal III and impurity E are associated through the first pair of canonical variates and have a significant correlation. Other correlations are relatively weak and have little explanatory effect on impurities. Therefore, only the first pair of canonical variates and r1 in this example are statistically significant for the crystal form-impurity correlation.

[0104] Example 3: Screening of the relatively most stable crystal form

[0105] Cefathiamidine impurity E has a correlation with crystal form Ia and crystal form III through the first pair of canonical variates, and impurity B has a correlation with crystal form II through the first pair of canonical variates. The canonical correlation coefficient r1 is greater than 0.50, indicating a significant correlation, that is, crystal form II is more likely to produce impurity B; crystal forms Ia and III are more likely to produce impurity E. Therefore, impurity B and impurity E can be used as key process pointers related to crystal forms. Crystal form I has no correlation with all impurities. Within the range of the apparent crystal forms of the existing samples, crystal form I can be considered as the relatively most stable crystal form.

[0106] The results of canonical correlation analysis were verified using the impurity distribution of each manufacturer. The mean values, batch probabilities, and crystal distributions of impurity B and impurity E of each manufacturer are shown in Table 10. The results show that impurity B and crystal II coexist, while impurity E coexists with crystal forms Ia and III, which is consistent with the results of canonical correlation analysis and the correlation relationship transmission route.

[0107] Table 10 Mean values, batch probabilities, and crystal distributions of impurity B and impurity E of each manufacturer

[0108]

[0109] The above has described the embodiments of the present invention in detail, but the content described is only the preferred embodiments of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. Method for analyzing the correlation between the apparent crystal form of a mixed crystal type of polymorphic drug and impurity formation, characterized in that The specific steps are as follows: S1: Use the apparent crystal form characterization model to detect the apparent crystal form of the target drug and obtain big data on apparent crystal forms; screen the common impurity categories of the target drug to obtain big data on the impurity detection results of the target drug. The method for establishing the apparent crystal form characterization model is as follows: S1.1: Screen typical samples, detect each typical sample by MDSC to obtain the apparent crystal form model of each typical sample; establish the spectral model of each typical sample by near-infrared spectroscopy; establish the connection between the apparent crystal form model and the spectral model corresponding to the same typical sample. S1.2: Perform near-infrared spectral characterization on the samples to be detected, and combine the corresponding relationship between the spectral model and the apparent crystal form model to obtain big data information on apparent crystal forms. S2: Correlate the apparent crystal form information with the impurity information, and use the canonical correlation analysis method to analyze the correlation coefficients between the apparent crystal form and the impurities, including the canonical correlation coefficients between pairs of canonical variables and the correlation coefficients between the original variables and the canonical variables. Among them, the pairs of canonical variables include impurity canonical variables and apparent crystal form canonical variables, which are in a paired relationship. The specific analysis steps include: S2.1: Based on the content of various impurities and the content of various apparent crystal forms as data, extract multiple pairs of canonical variables and calculate the correlation coefficients between each pair of canonical variables. S2.2: Calculate the conversion coefficients between various impurities and impurity canonical variables, and between various apparent crystal forms and apparent crystal form canonical variables through the canonical correlation analysis method; generate the relationship formula for each canonical variable. Taking the apparent crystal form factor as the independent variable X and the impurity factor as the dependent variable Y, according to the correlation relationship between the two sets of variables, obtain the canonical variables, substitute the apparent crystal form and the impurities, and then project the relationship between the apparent crystal form and the impurities onto the relationship between the pairs of canonical variables to obtain the conversion coefficients between the original variables and the canonical variables. The calculation formulas for the typical apparent crystal form variable U and the typical impurity variable V are respectively: ； ； where i = 1, 2, …, M, M is the number of common factors, j = 1, 2, …, N, N is the number of observed samples; S2.3: Test the canonical correlation coefficients of the generated pairs of canonical variables. Test the canonical correlation coefficients through the likelihood ratio method, and screen the pairs of canonical variables with p values less than 0.

05. S2.4: Construct the correlation coefficient matrices between the multiple pairs of canonical variables obtained by screening and the impurities and the apparent crystal forms respectively, and obtain the correlation coefficients between the impurities, each canonical variable and the apparent crystal form. S3: Determine the transfer relationship according to the correlation coefficients between the impurities, the pairs of canonical variables and the crystal form, and screen the correlation degree between the specific category of apparent crystal form and the specific category of impurities. Among them, the apparent crystal form is the characteristics of the population mixed crystal such as the single crystal composition and its relative relationship in the mixed crystal of the polymorphic drug.

2. The method for analyzing the correlation between the apparent crystal form of the mixed crystal type of polymorphic drugs and the formation of impurities according to claim 1, characterized in that: In step S2.4, calculate the correlation coefficient matrix between impurities-impurity canonical variables, the correlation coefficient matrix between apparent crystal forms-apparent crystal form canonical variables, the correlation coefficient matrix between apparent crystal forms-impurity canonical variables, and the correlation coefficient matrix between impurities-apparent crystal form canonical variables respectively.

3. The method for analyzing the correlation between the apparent crystal form of the mixed crystal type of polymorphic drug and the formation of impurities according to claim 1, characterized in that Step S3 includes: S3.1: Screen the correlation coefficients not less than 0.3 as alternative transfer relationships; S3.2: It is determined that the apparent crystal forms and impurities of the correlation coefficients of the same group of typical impurity variables that are all positively correlated or all negatively correlated are correlated.

4. The method for analyzing the correlation between the apparent crystal form of the mixed crystal type of polymorphic drugs and the formation of impurities according to claim 3, characterized in that: In step S3.1, the correlation coefficient matrices of the apparent crystal form-impurity typical variables and the impurity-apparent crystal form typical variables are screened, and the correlation coefficients not less than 0.3 are selected, and the corresponding impurities or apparent crystal forms are screened. In step S3.2, based on the correlation coefficients in the impurity-impurity typical variable correlation coefficient matrix and the apparent crystal form-apparent crystal form typical variable correlation coefficient matrix, it is compared whether the impurities and apparent crystal forms for the same group of typical variable pairs are all positively correlated or all negatively correlated.

5. The method for analyzing the correlation between the apparent crystal form of the mixed crystal type of polymorphic drugs and the formation of impurities according to claim 3, characterized in that: The correlation coefficients of the correlated impurities and apparent crystal forms are equal to the canonical correlation coefficients between the corresponding typical variable pairs.

6. The analysis method for the correlation between the apparent crystal form of the mixed crystal type of polymorphic drugs and impurities as described in any one of claims 1-5, characterized in that It also includes: S4: The apparent crystal form with no or low correlation with the impurity is screened as the most stable crystal form.

Images