CN114384040A - Method for establishing general test model of physical property indexes of solid preparation intermediate - Google Patents
Method for establishing general test model of physical property indexes of solid preparation intermediate Download PDFInfo
- Publication number
- CN114384040A CN114384040A CN202011131359.4A CN202011131359A CN114384040A CN 114384040 A CN114384040 A CN 114384040A CN 202011131359 A CN202011131359 A CN 202011131359A CN 114384040 A CN114384040 A CN 114384040A
- Authority
- CN
- China
- Prior art keywords
- model
- physical property
- modeling
- solid preparation
- pretreatment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 89
- 230000000704 physical effect Effects 0.000 title claims abstract description 42
- 239000007787 solid Substances 0.000 title claims abstract description 26
- 238000002360 preparation method Methods 0.000 title claims abstract description 21
- 238000012812 general test Methods 0.000 title claims abstract description 13
- 238000012216 screening Methods 0.000 claims abstract description 44
- 238000012795 verification Methods 0.000 claims abstract description 36
- 238000012937 correction Methods 0.000 claims abstract description 35
- 238000002329 infrared spectrum Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 28
- 238000001228 spectrum Methods 0.000 claims description 55
- 239000002245 particle Substances 0.000 claims description 53
- 239000008187 granular material Substances 0.000 claims description 49
- 239000002775 capsule Substances 0.000 claims description 30
- 238000009499 grossing Methods 0.000 claims description 23
- 239000002994 raw material Substances 0.000 claims description 22
- 239000000203 mixture Substances 0.000 claims description 7
- 238000009472 formulation Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 2
- 239000000543 intermediate Substances 0.000 abstract description 74
- 238000004458 analytical method Methods 0.000 abstract description 14
- 238000004519 manufacturing process Methods 0.000 abstract description 14
- 238000001514 detection method Methods 0.000 abstract description 10
- 230000003595 spectral effect Effects 0.000 abstract description 10
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 62
- 239000000843 powder Substances 0.000 description 42
- 238000011156 evaluation Methods 0.000 description 35
- 101150061025 rseP gene Proteins 0.000 description 23
- 238000002203 pretreatment Methods 0.000 description 20
- 101001078093 Homo sapiens Reticulocalbin-1 Proteins 0.000 description 16
- 244000131316 Panax pseudoginseng Species 0.000 description 16
- 235000003181 Panax pseudoginseng Nutrition 0.000 description 16
- 102100025335 Reticulocalbin-1 Human genes 0.000 description 16
- 230000036961 partial effect Effects 0.000 description 13
- 238000001035 drying Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 11
- 238000002156 mixing Methods 0.000 description 11
- 238000010606 normalization Methods 0.000 description 9
- 230000002829 reductive effect Effects 0.000 description 9
- 238000011160 research Methods 0.000 description 9
- 241000427159 Achyranthes Species 0.000 description 8
- 241001254604 Angelica pubescens Species 0.000 description 8
- 241000218176 Corydalis Species 0.000 description 8
- 238000004497 NIR spectroscopy Methods 0.000 description 8
- 235000006484 Paeonia officinalis Nutrition 0.000 description 8
- 241001106477 Paeoniaceae Species 0.000 description 8
- 241000219061 Rheum Species 0.000 description 8
- 235000009411 Rheum rhabarbarum Nutrition 0.000 description 8
- 239000000706 filtrate Substances 0.000 description 8
- 238000001914 filtration Methods 0.000 description 8
- 239000000463 material Substances 0.000 description 8
- 239000000047 product Substances 0.000 description 8
- 238000005070 sampling Methods 0.000 description 8
- 238000005303 weighing Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 239000012530 fluid Substances 0.000 description 7
- 238000002790 cross-validation Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 238000010200 validation analysis Methods 0.000 description 5
- 241000212322 Levisticum officinale Species 0.000 description 4
- 235000003143 Panax notoginseng Nutrition 0.000 description 4
- 241000180649 Panax notoginseng Species 0.000 description 4
- 239000000853 adhesive Substances 0.000 description 4
- 230000001070 adhesive effect Effects 0.000 description 4
- 239000003814 drug Substances 0.000 description 4
- 229940079593 drug Drugs 0.000 description 4
- 239000001645 levisticum officinale Substances 0.000 description 4
- 239000000314 lubricant Substances 0.000 description 4
- VYQNWZOUAUKGHI-UHFFFAOYSA-N monobenzone Chemical compound C1=CC(O)=CC=C1OCC1=CC=CC=C1 VYQNWZOUAUKGHI-UHFFFAOYSA-N 0.000 description 4
- 238000010298 pulverizing process Methods 0.000 description 4
- 238000007873 sieving Methods 0.000 description 4
- 238000002791 soaking Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000005550 wet granulation Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000007430 reference method Methods 0.000 description 3
- 238000000638 solvent extraction Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000011426 transformation method Methods 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000005034 decoration Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 239000012467 final product Substances 0.000 description 2
- 238000002955 isolation Methods 0.000 description 2
- 231100000862 numbness Toxicity 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000004611 spectroscopical analysis Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 239000003826 tablet Substances 0.000 description 2
- 229940126680 traditional chinese medicines Drugs 0.000 description 2
- 238000000149 argon plasma sintering Methods 0.000 description 1
- 235000009120 camo Nutrition 0.000 description 1
- 238000012569 chemometric method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000007865 diluting Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000007429 general method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000007427 paired t-test Methods 0.000 description 1
- 238000007790 scraping Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000012353 t test Methods 0.000 description 1
- 235000020985 whole grains Nutrition 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N1/00—Sampling; Preparing specimens for investigation
- G01N1/28—Preparing specimens for investigation including physical details of (bio-)chemical methods covered elsewhere, e.g. G01N33/50, C12Q
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention provides a method for establishing a general test model of physical property indexes of a solid preparation intermediate, which comprises the following steps: (1) selecting a solid preparation intermediate sample, dividing the intermediate sample into a correction set and a verification set, and collecting a near infrared spectrum; (2) determining a physical property indicator for the intermediate sample; (3) pre-processing the near infrared spectrum; (4) after the pretreatment, characteristic variable screening is carried out; (5) and establishing a general test model of the physical property index based on the pretreatment and the characteristic variable screening. The feasibility of establishing a near-infrared universal quantitative model for the intermediate in the production process of the variety is discussed through the relation analysis of near-infrared (NIR) spectral information and physical property indexes, and the result shows that the near-infrared universal quantitative model can be used for measuring various intermediates, so that the detection efficiency is greatly improved.
Description
Technical Field
The invention relates to the field of solid preparation detection, in particular to a method for establishing a general test model of physical property indexes of a solid preparation intermediate.
Background
In the solid formulation forming process stage, the physical properties of the intermediate become a major factor affecting the quality of the final product. The physical properties of the particles also have an impact on a number of key quality attributes, such as flowability, compressibility, and thus the quality of the final product, and the efficacy of the formulated product. At present, in the process of optimizing and improving the quality standard of a plurality of solid preparations, some physical property indexes are brought into the quality internal control standard.
Near Infrared (NIR) spectroscopy is combined with a chemometric method, so that the content of chemical components and physical property indexes can be rapidly detected, and the method is more applied to medicinal material identification and rapid determination of the content of the chemical components, but less used for rapidly detecting the physical property indexes.
Most studies of near infrared spectroscopy combined with stoichiometry have been directed to the analysis of a certain class of samples. Establishing a stable and reliable NIR quantitative prediction model is a relatively complicated task, long in time and high in cost. If a universal model can be established, the detection efficiency can be greatly improved, and the cost is saved.
Disclosure of Invention
In view of the above, one of the main objectives of the present invention is to provide a general test model for physical property indexes of solid preparation intermediates by using near infrared spectroscopy technology, and a method for establishing the model, so as to realize rapid detection of physical property indexes of particles.
In order to achieve the above object, the present invention provides a method for establishing a general test model for physical property indexes of an intermediate of a solid preparation, the method comprising:
(1) selecting a solid preparation intermediate sample, dividing the intermediate sample into a correction set and a verification set, and collecting a near infrared spectrum;
(2) determining physical property indicators for the solid formulation intermediate sample, wherein physical property indicators include, but are not limited to, particle size, angle of repose, hardness, hygroscopicity, bulk density;
(3) screening out a proper pretreatment mode from a near infrared spectrum pretreatment method, thereby pretreating the near infrared spectrum based on the physical property index; wherein, the preprocessing methods include but are not limited to moving window smoothing (9 points), SNV, S-G1 st (9 points), baseline correction, MSC, normalization;
(4) after the pretreatment, screening characteristic variables by adopting different modes, such as a mode based on an interval partial least square method, a combined interval partial least square method, a moving window partial least square method and the like, and comprehensively evaluating and screening a modeling waveband by taking RMSEC, RMSECV, RMSEP and the like as evaluation indexes;
(5) and determining the number of main factors by taking RMSECV as an evaluation index and establishing a general test model of the physical property index based on the pretreatment and the characteristic variable screening result, wherein the general test model is used for detecting the intermediate physical property index.
The general model is a common model established for two or more samples to realize the rapid detection of the relevant indexes of various samples.
Further, the intermediate includes at least two of raw material fine powder, dry granules, whole granules, and total mixed granules.
Optionally, the solid preparation comprises capsules, tablets and granules.
Specifically, the capsule can be a waist numbness relieving capsule.
Further, the intermediate physical property index can be the particle size of the lubelongtong capsule intermediate, preferably the median particle size (D)50) Hygroscopicity and bulk density.
Further, the pre-processing may be selected from raw spectra or SNV pre-processing or first derivative combined SG smoothing pre-processing modeling.
Further, the feature variable screening may be selected from full band modeling or preferably optimal spectral interval modeling.
The invention also provides a method for detecting physical property indexes of the intermediate of the solid preparation, which comprises the following steps:
(1) selecting a solid preparation intermediate sample, dividing the intermediate sample into a correction set and a verification set, and collecting a near infrared spectrum;
(2) determining physical property indicators for the solid formulation intermediate sample, wherein physical property indicators include, but are not limited to, particle size, bulk density, angle of repose, hardness, hygroscopicity, bulk density;
(3) screening out a proper pretreatment mode from a near infrared spectrum pretreatment method, thereby pretreating the near infrared spectrum based on the physical property index;
(4) after the pretreatment, characteristic variable screening is carried out;
(5) determining the number of main factors by taking RMSECV as an evaluation index and establishing a general test model of physical property indexes on the basis of the pretreatment and the characteristic variable screening result;
(6) and detecting the physical property index of the intermediate to be detected by using the universal test model.
Further, the intermediate includes one or at least two of the raw material fine powder, the dry granules, the whole granules and the total mixed granules, and preferably the intermediate includes four of the raw material fine powder, the dry granules, the whole granules and the total mixed granules.
Alternatively, the solid formulation may comprise capsules, tablets, granules.
Specifically, the capsule can be a waist numbness relieving capsule.
The optional preparation method of the lumbar rheumatism capsule comprises the following steps:
(1) 100 portions of pseudo-ginseng, 500 portions of szechuan lovage rhizome, 200 portions of rhizoma corydalis, 500 portions of white paeony root, 200 portions of white paeony root, 500 portions of twotooth achyranthes root, 200 portions of twotooth achyranthes root, 400 portions of rhizoma cibotii, 400 portions of prepared rhubarb, 100 portions of pubescent angelica root, 400 portions of pubescent angelica root, and the eight medicines are obtained by crushing half of the pseudo-ginseng into fine powder and the other half of the pseudo-ginseng into coarse powder;
(2) pulverizing rhizoma corydalis, rhizoma Ligustici Chuanxiong, and radix Angelicae Pubescentis into coarse powder, mixing with Notoginseng radix coarse powder, percolating with 4-8 times of 65-85% ethanol, and making into fluid extract;
(3) soaking radix paeoniae alba, radix achyranthis bidentatae, rhizoma cibotii and cooked rhubarb in water, decocting for 1-2 times, each time for 30-90 minutes, concentrating the extracting solution until the relative density is 1.15-1.20 at 50 ℃, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, concentrating until the relative density is 1.15-1.20, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, and concentrating to a proper volume;
(4) adding the product prepared in the step (3) into the product prepared in the step (2), diluting, adding the fine powder of the pseudo-ginseng to prepare a mixed extract, drying, and crushing into raw material fine powder.
(5) And (4) adding a proper amount of adhesive into the fine powder of the raw material prepared in the step (4), performing wet granulation, drying and crushing to obtain dry granules.
(6) And (4) adding a proper amount of lubricant into the dried granules prepared in the step (5), and sieving to prepare the whole granules.
(7) And (4) uniformly mixing the whole granules prepared in the step (6) to obtain total mixed granules.
Further, the intermediate physical property index may be a particle size of the lubelongtong capsule intermediate, preferably a median particle size (D)50) Hygroscopicity and bulk density.
Specifically, the near infrared spectrum pretreatment method can comprise the following steps: smoothing methods (carriage average method, moving window smoothing method, convolution smoothing method, Norris smoothing method), non-preprocessing (original spectrum), derivative methods (first derivative method, second derivative method), vector normalization method, light scattering correction method, standard normal variable transformation method, orthogonal signal correction method, baseline correction method, wavelet transformation method, fourier transformation method, net analysis signal method, and the like.
Specifically, the characteristic variable includes a full band or a band after being screened by a screening method. The characteristic variable screening mode comprises an interval partial least square method, a combined interval partial least square method, a backward interval partial least square method, a moving window partial least square method, a continuous projection method, a competitive self-adaptive weighted sampling method, a genetic algorithm, a particle swarm optimization method, an information-free variable elimination method, a simulated annealing algorithm and the like.
Further, the preprocessing method is selected from raw spectrum or SNV preprocessing or modeling of first derivative combined with SG smoothing preprocessing; and the characteristic variable screening is full-band modeling or optimal spectrum interval modeling.
The invention discusses the feasibility of establishing a near infrared universal quantitative model for the intermediate in the production process of the variety through the relation analysis of Near Infrared (NIR) spectral information and physical property indexes. The general model can be used for quantitative determination, and the prediction accuracy is higher. Therefore, the near-infrared universal quantitative model of the invention can be used for physical property indexes such as D of various intermediates of the capsule such as the Yabitong capsule50And the moisture absorption and the bulk density are measured, so that the detection efficiency is greatly improved, and the detection cost is saved.
Drawings
FIG. 1 is a near-infrared original spectrum of a Yaotong capsule intermediate;
FIG. 2 is D50Correlation of the predicted value with the reference value; wherein, A, a raw material fine powder model; B. drying the particle model; C. a whole granule model; D. a total mixed particle model; E. a general model;
FIG. 3 is a correlation of hygroscopicity reference values with predicted values;
FIG. 4 is a graph of the number of principal factors of hygroscopicity as a function of RMSECV;
FIG. 5 is a correlation of a bulk density reference value with a predicted value.
Detailed Description
One of the main purposes of the invention is to provide a general test model of the physical properties of the intermediate of the solid preparation by utilizing the near infrared spectrum technology and a method for establishing the model. The following detailed description of the present invention, taken in conjunction with the accompanying drawings and examples, is provided to enable the invention and its various aspects and advantages to be better understood. However, the specific embodiments and examples described below are for illustrative purposes only and are not limiting of the invention.
It is specifically noted that similar alternatives and modifications will be apparent to those skilled in the art, which are also intended to be included within the present invention. It will be apparent to those skilled in the art that the techniques of the present invention may be implemented and applied by modifying or appropriately combining the methods and applications described herein without departing from the spirit, scope, and content of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. In addition, any methods and materials similar or equivalent to those described herein can be used in the practice of the present invention.
In the embodiment, the capsule for treating lumbar rheumatism is taken as an example, an intermediate in the production process is taken as a research object, and the intermediate D is taken as an intermediate D by collecting the near-infrared diffuse reflection spectrum of the intermediate50The hygroscopicity and the bulk density are taken as reference values, and a plurality of intermediates D for detection are established50General methods of hygroscopicity and bulk density.
Near-infrared universal quantitative model research on intermediate particle size of lumbar arthromyodynia capsule
1 instruments and materials
1.1 instruments
An Antaris II type Fourier near infrared transform spectrometer which is provided with an integrating sphere diffuse reflection sampling system and Result spectrum acquisition software, and is manufactured by Thermo company of America; BT-2600 laser particle size distribution Instrument, Dandong Baite instruments, Inc.
1.2 materials
The intermediates in the production process of the Yabitong capsule comprise 320 raw material fine powder, dry granules, whole granules and total mixed granules, and are provided by Jiangsu Kangyuan pharmaceutical industry GmbH. The production process of the lumbar rheumatism capsule in the embodiment is as follows:
(1) 100 portions of pseudo-ginseng, 500 portions of szechuan lovage rhizome, 200 portions of rhizoma corydalis, 500 portions of white paeony root, 200 portions of white paeony root, 500 portions of twotooth achyranthes root, 200 portions of twotooth achyranthes root, 400 portions of rhizoma cibotii, 400 portions of prepared rhubarb, 100 portions of pubescent angelica root, 400 portions of pubescent angelica root, and the eight medicines are obtained by crushing half of the pseudo-ginseng into fine powder and the other half of the pseudo-ginseng into coarse powder;
(2) pulverizing rhizoma corydalis, rhizoma Ligustici Chuanxiong, and radix Angelicae Pubescentis into coarse powder, mixing with Notoginseng radix coarse powder, percolating with 4-8 times of 65-85% ethanol, and making into fluid extract;
(3) soaking radix paeoniae alba, radix achyranthis bidentatae, rhizoma cibotii and cooked rhubarb in water, decocting for 1-2 times, each time for 30-90 minutes, concentrating the extracting solution until the relative density is 1.15-1.20 at 50 ℃, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, concentrating until the relative density is 1.15-1.20, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, and concentrating to a proper volume;
(4) adding the fluid extract obtained in the step (2) into the product obtained in the step (3), recovering ethanol, concentrating into a thick paste, drying, adding the fine powder of the pseudo-ginseng, uniformly mixing to obtain a mixed extract, drying, and crushing into raw material fine powder.
(5) And (4) adding a proper amount of adhesive into the fine powder of the raw material prepared in the step (4), performing wet granulation, drying and crushing to obtain dry granules.
(6) And (4) adding a proper amount of lubricant into the dried granules prepared in the step (5), and sieving to prepare the whole granules.
(7) And (4) uniformly mixing the whole granules prepared in the step (6) to obtain total mixed granules.
2 method
2.1 near Infrared Spectroscopy
Taking about 4g of sample, placing the sample in a prepared sample cup,lightly compacting, and collecting near infrared spectrum by integrating sphere diffuse reflection. The scanning range is 10000-4000 cm-1Resolution of 8cm-1Gain 2, scan 64 times background air, once per hour. Each sample was scanned 3 times and the average was used for analysis. The near infrared original spectrum of the lubelongtong capsule intermediate is shown in figure 1.
2.2 reference value determination
Measuring the particle size distribution of the sample by using a laser particle size distribution instrument, and calculating D50(cumulative distribution up to 50% corresponds to particle size).
2.3 Spectrum pretreatment method
When the spectrum is collected, due to the influence of various factors such as environment temperature and humidity, instrument state, particle state and the like, the near infrared spectrum can generate noise signals, baseline drift and the like, and the acquired spectrum contains self information and other unnecessary signals. The near infrared spectrum is subjected to proper pretreatment, so that noise can be reduced, irrelevant information can be filtered, and the robustness of the model can be improved. Such as vector normalization (normalization), Derivative (first Derivative, second Derivative), smoothing [ Savitzky-golay (sg) smoothing, Norris Derivative (ND) smoothing ], standard normal variable transformation (SNV), Multiple Scatter Correction (MSC), baseline correction, and combinations thereof.
2.4 data processing and evaluation method
NIR spectra were pre-processed using Unscramble X10.4 (Camo software AS, Norway) software and samples were variable screened and modeled using Matlab 2016a (Mathwork inc., USA) software. The near infrared quantitative model generally takes the following parameters as model performance evaluation indexes: a performance deviation Ratio (RPD), a correction set correlation coefficient (Rcal), a verification set correlation coefficient (Rpre), a corrected Root Mean Square Error (RMSEC), a predicted Root Mean Square Error (RMSEP), a predicted relative deviation (RSEP), a cross-verified root mean square error (rmcv), a corrected set deviation (biasca), and a verified set deviation (biasrre). The RPD represents the prediction performance of the model, and when the RPD is more than 3, the prediction accuracy of the model is high. The larger Rcal and Rpre, the smaller RMSECV, RMSEC, RMSEP, RSEP, BIAScal and BIASPRE, the better the model correction and prediction performance.
3 results and analysis
3.1 sample partitioning
Dividing a correction set and a verification set by each intermediate according to the ratio of 3:1, randomly selecting 60 samples as the correction set, and 20 samples as the verification set; the correction set of the universal model is the sum of the four intermediate correction sets, and the total number of the samples is 240, and the verification set is the sum of the four intermediate verification sets, and the total number of the samples is 80.
TABLE 1 statistical results of the reference method measurements of the calibration and validation sets of samples
3.2 selection of spectral Pre-processing method
This embodiment considers the following preprocessing methods: moving window smoothing (9 points), SNV, S-G1 st (9 points), baseline correction, MSC, normalization, etc. Spectra were pre-treated using the above method, and the effect of different pre-treatment methods on model performance is shown in table 2. And (3) screening a pretreatment method by taking RPD and RSEP as evaluation standards. As can be seen from Table 2, in the raw material fine powder model, the effect of pre-processing and modeling the spectrum by combining first-order derivation with S-G smoothing is optimal, the RPD is 4.50, and the RSEP is 2.30%; in the dry particle model, the best pretreatment method is adopted by moving window smoothing (9 points), the RPD is 4.12, and the RSEP is 1.88%; in the whole grain particle model, the original spectrum modeling ratio is adopted for carrying out pretreatment, the modeling effect is good, the RPD is 4.15, and the RSEP is 1.80%; the total mixed particle model has the best modeling effect after preprocessing the spectrum by adopting a moving window smoothing (9 points) method, wherein the RPD is 4.84, and the RSEP is 1.54%; in the general model, the original spectrum is adopted to model the best performance, the RPD is 4.60, and the RSEP is 2.26%.
TABLE 2 Effect of different pretreatment methods on the model
3.3 feature variable screening
And the screening of the characteristic variables can eliminate irrelevant information and improve the performance of the model. The present embodiment further screens characteristic variables on the basis of the above-described screened preprocessing method.
3.3.1 screening variables based on Interval partial least squares (iPLS)
iPLS is a method in which the full spectrum is divided into several sub-intervals, and then modeling is performed in each sub-interval. In the embodiment, the full spectrum is divided into 20 subintervals, and the RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.2 screening variables based on Combined Interval partial least squares (sPLS)
The siPLS is based on iPLS, and the full spectrum is divided into a plurality of subintervals, and then the subintervals are randomly combined for modeling. In the embodiment, the full spectrum is divided into 20 subintervals, the number of combinations of the subintervals is 4, a model is established, and RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.3 screening of variables based on moving Window PLS (mwPLS)
The mwPLS is a series of PLS models created by taking intervals of selected window widths in the direction of wavelength change, starting from the first wavelength point of the entire spectrum. In the embodiment, the initial window width is 31, the window widths are sequentially increased by taking 10 as a step length, a PLS model with the window width of 31-311 is established, and an optimal modeling waveband is selected according to RMSECV.
3.3.4 screening for characteristic variables
The performance parameters of the model established by the 3 methods are shown in tables 3-7. And comprehensively evaluating and screening the optimal modeling waveband by taking RMSEC, RMSECV and RMSEP as evaluation indexes.
The performance of a raw material fine powder model is not greatly different from that of a mwPLS preferred waveband modeling, but the variables are reduced to 115 after the waveband is screened, and the modeling time is shortened, so that the modeling waveband is selected to be 3999.64-4018.92 cm-1、5739.12～6136.38cm-1。
After the dry particle model adopts the siPLS and mwPLS to screen variables, the model performance is improved, the variable numbers are respectively reduced to 392 and 278, but the mwPLS screens the variables, the wave point number is less, the Rcal and the Rpre are larger, the RMSEC and the RMSEP are smaller, and therefore, the wave band is 4030.50-4219.49 cm-1、4443.19～4504.90cm-1、6957.91～7767.86cm-1And (6) modeling.
The performance of a variable model screened by a sipPLS and a mwPLS is improved, the number of variables is reduced to 312 and 83 respectively, but the Rcal and Rpre of the mwPLS are larger, the RMSEC and RMSEP are smaller, and the number of wave points is less, so that the wave band is 4933.02-4971.59 cm-1、5982.10～6213.52cm-1、7359.03～7397.60cm-1And (6) modeling.
After the total mixed particle model is subjected to wave band screening by adopting the mwPLS method, the Rcal and the Rpre are both increased, the RMSEC and the RMSEP are both decreased, and the performance of the model is improved, so that the selected wave band is 4015.07-4242.63 cm-1、4396.90～4524.18cm-1、7667.58～7698.44cm-1、8369.54～8454.40cm-1And (6) modeling.
Although the complexity of the model is reduced after the universal model screens variables, the model performance is inferior to that of a full spectrum model, probably because the universal model is a common model containing four intermediate information, important information may be lost by adopting band screening, so that the performance of the model is reduced, and the universal model adopts full-band modeling.
TABLE 3 comparison of raw material fines optimization interval modeling with full spectrum modeling
TABLE 4 comparison of Dry particle preferred Interval modeling with full Spectroscopy modeling
TABLE 5 comparison of Whole particle preferred Interval modeling with full Spectroscopy modeling
TABLE 6 comparison of Total Mixed particle preferred Interval modeling with full Spectrum modeling
TABLE 7 comparison of the Universal model preferred Interval modeling with full Spectrum modeling
3.4 determination of the number of major factors
The number of main factors is determined by taking the RMSECV value as an evaluation index, the improper selection of the number of main factors can cause insufficient fitting or excessive fitting, and the corresponding number of main factors is generally the best when the RMSECV value is the minimum. In the embodiment, a leave-one-out cross verification method is adopted, RMSECV is used as an index, and the influence of the number of main factors on the model is examined. Finally, the optimal main factor numbers of the raw material fine powder model, the dry particle model, the whole particle model, the total mixed particle model and the universal model are respectively determined to be 6, 15, 12, 11 and 14.
3.5 D50Establishment of quantitative prediction model
Pretreatment method adopting screening and NIR spectral band establishment D50The near infrared quantitative prediction model is shown in figure 2. The evaluation parameters of the raw material fine powder, the dried particles, the whole particles, the total mixed particles and the general model Rcal are 0.941, 0.966, 0.976, 0.977 and 0.965 in sequence, Rpre is 0.975, 0.979, 0.975, 0.980 and 0.977 in sequence, and RMSEC is 4.010 μm, 2.261 μm, 1.905 μm, 2.073 μm and 3.392 μm in sequence; RMSECV is 4.934 μm, 3.603 μm, 3.158 μm, 2.692 μm and 3.918 μm in sequence; RMSEP is 2.542 μm, 2.214 μm, 2.163 μm, 1.870 μm and 2.832 μm in sequence, and RSEP is 2.29%, 1.70%, 1.66%, 1.45% and 2.26% in sequence; the RPD is 4.51, 4.55, 4.49, 5.13 and 4.60 in sequence. Rcal and Rpre are close to 1, which shows that the correlation between the reference value and the predicted value is high; the RMSEC, RMSECV, RMSEP and RSEP are small, the RPD is more than 3, which shows that the model has higher prediction performance and can be used for D50Quantitative prediction of (3).
3.6 model verification and evaluation
Introducing NIR spectra of verification set samples into a correction model to predict D of intermediate50And compared with the reference value, the results are shown in tables 8-9. Raw material fine powder model, dry particle model, whole particle model, total mixed particle model and general model D50The average relative prediction errors of the two models are respectively 1.94%, 1.43%, 1.46%, 1.21% and 1.87%, and are all less than 5%, which indicates that the prediction accuracy of the two models is higher. As can be seen from Table 10, the independent model and the general model were applied to each intermediate D50The difference of the prediction errors is less than 1%, which indicates that the prediction performances of the two are similar.
And (3) carrying out pairing t test on the reference values and the predicted values in the verification set of the raw material fine powder, the dry granules, the whole granules, the total mixed granules and the universal model, wherein the t values are 0.22, 0.17, -0.42, -0.38 and 0.75 in sequence, the P values are 0.826, 0.865, 0.678, 0.707 and 0.454 in sequence, and the P values are all more than 0.05, which indicates that no significant difference exists between the reference values and the predicted values.
TABLE 8 comparison of near Infrared reference values with predicted values (independent model)
TABLE 9 near Infrared reference vs. predicted values (general model)
TABLE 10 independent model versus generic model relative prediction error comparison
In the embodiment, the Yabi Tong capsule is taken as a research object, 4 intermediates including raw material fine powder, dry particles, whole particles and total mixed particles in the production process are collected, the intermediate differences are small, and the near infrared spectrum technology is adopted to establish the prediction intermediate D50The universal quantitative model of (1). All evaluation indexes show that the universal quantitative model has high prediction accuracy, can meet the actual requirements, has little difference with a single model in the prediction performance, and can realize the prediction of 4 intermediates D50The measurement of (1).
Compared with a laser particle size distribution instrument method, the particle size can be measured more quickly by using a near infrared spectroscopy, and the method is applied to on-line research without sampling, so that the analysis efficiency is greatly improved. Because the near-infrared general model is less researched in the pharmaceutical field, the near-infrared spectrum technology is combined with the general model, the established near-infrared general model has the advantages of rapidness, accuracy, no damage, greenness, cost saving and the like, and a better method can be provided for realizing online detection of related indexes in the production process of traditional Chinese medicines.
Near-infrared universal quantitative model research on hygroscopicity of lumbar arthromyodynia capsule intermediate
1 instruments and materials
1.1 instruments
An Antaris II type Fourier near infrared transform spectrometer which is provided with an integrating sphere diffuse reflection sampling system, Result spectrum acquisition software and TQ analysis 9.0 analysis software, and is manufactured by the American Thermo company; constant temperature and humidity chamber, shanghai-chang scientific instruments ltd.
1.2 materials
120 intermediates in the production process of the Yabitong capsule, which comprise dry particles and total mixed particles, are provided by Jiangsu Kangyuan pharmaceutical industry GmbH. The production process of the lumbar rheumatism capsule in the embodiment is as follows:
(1) 100 portions of pseudo-ginseng, 500 portions of szechuan lovage rhizome, 200 portions of rhizoma corydalis, 500 portions of white paeony root, 200 portions of white paeony root, 500 portions of twotooth achyranthes root, 200 portions of twotooth achyranthes root, 400 portions of rhizoma cibotii, 400 portions of prepared rhubarb, 100 portions of pubescent angelica root, 400 portions of pubescent angelica root, and the eight medicines are obtained by crushing half of the pseudo-ginseng into fine powder and the other half of the pseudo-ginseng into coarse powder;
(2) pulverizing rhizoma corydalis, rhizoma Ligustici Chuanxiong, and radix Angelicae Pubescentis into coarse powder, mixing with Notoginseng radix coarse powder, percolating with 4-8 times of 65-85% ethanol, and making into fluid extract;
(3) soaking radix paeoniae alba, radix achyranthis bidentatae, rhizoma cibotii and cooked rhubarb in water, decocting for 1-2 times, each time for 30-90 minutes, concentrating the extracting solution until the relative density is 1.15-1.20 at 50 ℃, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, concentrating until the relative density is 1.15-1.20, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, and concentrating to a proper volume;
(4) adding the fluid extract obtained in the step (2) into the product obtained in the step (3), recovering ethanol, concentrating into a thick paste, drying, adding the fine powder of the pseudo-ginseng, uniformly mixing to obtain a mixed extract, drying, and crushing into raw material fine powder.
(5) And (4) adding a proper amount of adhesive into the fine powder of the raw material prepared in the step (4), performing wet granulation, drying and crushing to obtain dry granules.
(6) And (4) adding a proper amount of lubricant into the dried granules prepared in the step (5), and sieving to prepare the whole granules.
(7) And (4) uniformly mixing the whole granules prepared in the step (6) to obtain total mixed granules.
2 method
2.1 near Infrared Spectroscopy
And (3) placing about 4g of sample in a prepared sample cup, lightly compacting, and collecting the near infrared spectrum by adopting an integrating sphere diffuse reflection mode. The scanning range is 10000-4000 cm-1Resolution of 8cm-1Gain 2, scan 64 times background air, once per hour. Each sample was scanned 3 times and the average was used for analysis.
2.2 reference value determination
Placing the dried bottle with stopper in a constant temperature and humidity chamber (set conditions are, temperature is 25 deg.C, relative humidity is 75%), precisely weighing after 12h (m1). Taking a proper amount of the intermediate sample, placing the intermediate sample in the weighing bottle, flatly paving the intermediate sample to obtain a product with the thickness of about 1mm, and precisely weighing the product (m)2). The weighing bottle is opened, and the bottle cap is placed under the constant temperature and humidity condition. After 24h, the weighing bottle is covered, and precision weighing is carried out (m)3) In parallel, 3 trials were performed.
2.3 Spectrum pretreatment method
When the spectrum is collected, the near infrared spectrum contains redundant signals due to the influence of various factors such as environment temperature and humidity, instrument state, particle state and the like. The near infrared spectrum is subjected to proper pretreatment, so that noise can be reduced, irrelevant information can be filtered, and the robustness of the model can be improved. Such as vector normalization (normalization), Derivative (first Derivative, second Derivative), smoothing [ Savitzky-golay (sg) smoothing, Norris Derivative (ND) smoothing ], standard normal variable transformation (SNV), etc.
2.4 data processing and evaluation method
NIR spectra were pre-processed and modeled using TQ Analyst 9.0 analysis software (Thermo, USA) and samples were variable screened using Matlab 2016a (Mathwork inc., USA) software. And determining the number of main factors by using a leave-one-out cross validation method and using a cross validation Root Mean Square Error (RMSECV) as an evaluation index. And (3) establishing a PLS quantitative model by taking the hygroscopicity of the intermediate as a dependent variable and the corresponding near infrared spectrum value as an independent variable. The present embodiment evaluates the model performance with the following evaluation parameters to optimize the optimal model. The correlation coefficient (Rcal) of the correction set and the correlation coefficient (Rpre) of the verification set respectively represent the fitting degree of the correction model and the verification model, and the larger the Rcal and the Rpre are, the better the model fitting effect is; root Mean Square Errors (RMSEC) and Root Mean Square Errors (RMSEP) of prediction refer to root mean square errors between reference values and predicted values in the correction model and the verification model respectively, and the smaller the model is, the higher the prediction performance is; the correction set deviation (BIAScal) and the verification set deviation (BIASpre) respectively represent the deviation between the reference value and the predicted value in the correction model and the verification model, and the smaller the model is, the higher the prediction accuracy is. The predicted relative error (RSEP) is the relative error between the reference value and the predicted value in the model, and generally, the smaller the RSEP, the better the model prediction performance, and each evaluation parameter cannot be referred to in isolation and needs to be evaluated comprehensively.
3 results and analysis
3.1 sample partitioning
And dividing the correction set and the verification set by adopting a Random Sampling (RS) method, and ensuring that the reference value range in the verification set is contained in the correction set. In the embodiment, 48 samples are randomly selected from each intermediate, mixed to serve as a correction set of the general model, the remaining 24 samples serve as a verification set, and the division results are shown in table 11.
TABLE 11 statistical results of the values measured in the sample calibration set and the validation set reference methods
3.2 selection of spectral Pre-processing method
This embodiment considers the following pretreatment methods for dry granules, total mix granules (total 120 samples): no pretreatment, SNV, S-G1 st (9 points), ND 1 st. Spectra were pre-treated using the above method, and the effect of different pre-treatment methods on model performance is shown in table 12. The pretreatment method was screened using RSEP as an evaluation criterion. As can be seen from Table 12, the spectral modeling performance was best when SNV was used for pretreatment, with RSEP of 2.42%.
TABLE 12 Effect of different preprocessing methods on the model
3.3 feature variable screening
And the screening of the characteristic variables can eliminate irrelevant information and improve the performance of the model. The present embodiment further screens characteristic variables on the basis of the above-described screened preprocessing method.
3.3.1 screening variables based on Interval partial least squares (iPLS)
iPLS is a method in which the full spectrum is divided into several sub-intervals, and then modeling is performed in each sub-interval. In the embodiment, the full spectrum is divided into 20 subintervals, and the RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.2 screening variables based on Combined Interval partial least squares (sPLS)
The siPLS is based on iPLS, and the full spectrum is divided into a plurality of subintervals, and then the subintervals are randomly combined for modeling. In the embodiment, the full spectrum is divided into 20 subintervals, the number of combinations of the subintervals is 4, a model is established, and RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.3 screening of variables based on moving Window PLS (mwPLS)
The mwPLS is a series of PLS models created by taking intervals of selected window widths in the direction of wavelength change, starting from the first wavelength point of the entire spectrum. In the embodiment, the initial window width is 71, the window widths are sequentially increased by taking 10 as a step length, a PLS model with the window width of 71-201 is established, and an optimal modeling waveband is selected according to RMSECV.
3.3.4 screening for characteristic variables
The performance parameters of the model built using the 3 methods are shown in table 13. And comprehensively evaluating and screening the optimal modeling waveband by taking RMSEC, RMSECV and RMSEP as evaluation indexes. As can be seen from Table 13, after the band screening is performed by the siPLS method, the Rcal and the Rpre become large, the RSEP becomes small, the model performance is improved, the number of variables is reduced to 312, and the modeling efficiency is greatly improved, so that the band 5499.99-5203.00 cm is selected-1、6703.35～6406.37cm-1、7305.00～7008.05cm-1、9110.08～8813.09cm-1And (6) modeling.
TABLE 13 comparison of preferred Interval modeling with full Spectrum modeling
3.4 determination of the number of major factors
Improper selection of the number of primary factors results in an inadequate or excessive fit, and is generally optimized for the number of primary factors that correspond to the smallest RMSECV value. In the present embodiment, a leave-one-out cross-validation method is used, and the influence of the number of main factors on the model is examined using RMSECV as an index, as shown in fig. 3. As can be seen from fig. 3, when the number of main factors is 7, the corresponding RMSECV is minimum, and therefore the optimal number of main factors for the model is determined to be 7.
3.5 general quantitative prediction model for hygroscopicity
A general quantitative prediction model for the hygroscopicity of the intermediate is established by adopting the screened pretreatment method and the NIR spectral band, and is shown in figure 4. The evaluation parameters of the model, Rcal is 0.912, Rpre is 0.945 and RMSEC is 0.486%; RMSECV 0.545%; RMSEP was 0.441%, and RSEP was 2.13%. Rcal and Rpre are close to 1, which shows that the correlation between the reference value and the predicted value is high; the smaller RMSEC, RMSECV, RMSEP and RSEP shows that the model has higher prediction performance and can be used for quantitative prediction of hygroscopicity.
3.6 model verification and evaluation
The NIR spectra of the validation set samples were introduced into a calibration model to predict the hygroscopicity of the intermediates and compared to reference values, the results of which are shown in table 14. The average relative prediction error of the hygroscopicity of the general model is 1.69 percent and is less than 5 percent, which indicates that the prediction accuracy of the model is higher. And carrying out paired t-test on the reference value and the predicted value in the verification set of the model, wherein the t value is-1.33, the P value is 0.198, and the P value is more than 0.05, which indicates that no significant difference exists between the reference value and the predicted value.
TABLE 14 near-IR reference value vs. predicted value
In the embodiment, two intermediates of dry particles and total mixed particles in the production process of the Yabitong capsule are taken as research objects, and a near infrared spectrum technology is adopted to establish a universal quantitative model for predicting the hygroscopicity of the intermediate. Various evaluation indexes show that the universal quantitative model has high prediction accuracy, can meet the actual requirement and can realize the determination of the hygroscopicity of two intermediates. Compared with a constant temperature and humidity box method, the near infrared spectroscopy can be used for measuring the moisture absorption more quickly without damage, and the analysis efficiency is greatly improved without sampling if the method is applied to on-line research. Because the near-infrared general model is less researched in the pharmaceutical field, the near-infrared spectrum technology is combined with the general model, the established near-infrared general model has the advantages of rapidness, accuracy, no damage, greenness, cost saving and the like, and a better method can be provided for realizing online detection of related indexes in the production process of traditional Chinese medicines.
Near-infrared universal quantitative model research on bulk density of lumbar arthromyodynia capsule intermediate
1 instruments and materials
1.1 instruments
An Antaris II type Fourier near infrared transform spectrometer which is provided with an integrating sphere diffuse reflection sampling system, Result spectrum acquisition software and TQ analysis 9.0 analysis software, and is manufactured by the American Thermo company; BT-1001 Intelligent powder Property tester, Dandongbeit instruments Co.
1.2 materials
120 intermediates in the production process of the Yabitong capsule are taken, and the intermediates comprise dry particles and whole particles, and are provided by Jiangsu Kangyuan pharmaceutical industry GmbH. The production process of the lumbar rheumatism capsule in the embodiment is as follows:
(1) 100 portions of pseudo-ginseng, 500 portions of szechuan lovage rhizome, 200 portions of rhizoma corydalis, 500 portions of white paeony root, 200 portions of white paeony root, 500 portions of twotooth achyranthes root, 200 portions of twotooth achyranthes root, 400 portions of rhizoma cibotii, 400 portions of prepared rhubarb, 100 portions of pubescent angelica root, 400 portions of pubescent angelica root, and the eight medicines are obtained by crushing half of the pseudo-ginseng into fine powder and the other half of the pseudo-ginseng into coarse powder;
(2) pulverizing rhizoma corydalis, rhizoma Ligustici Chuanxiong, and radix Angelicae Pubescentis into coarse powder, mixing with Notoginseng radix coarse powder, percolating with 4-8 times of 65-85% ethanol, and making into fluid extract;
(3) soaking radix paeoniae alba, radix achyranthis bidentatae, rhizoma cibotii and cooked rhubarb in water, decocting for 1-2 times, each time for 30-90 minutes, concentrating the extracting solution until the relative density is 1.15-1.20 at 50 ℃, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, concentrating until the relative density is 1.15-1.20, adding ethanol to ensure that the ethanol content reaches 60-90%, refrigerating for more than 24 hours, filtering, recovering ethanol from the filtrate, and concentrating to a proper volume;
(4) adding the fluid extract obtained in the step (2) into the product obtained in the step (3), recovering ethanol, concentrating into a thick paste, drying, adding the fine powder of the pseudo-ginseng, uniformly mixing to obtain a mixed extract, drying, and crushing into raw material fine powder.
(5) And (4) adding a proper amount of adhesive into the fine powder of the raw material prepared in the step (4), performing wet granulation, drying and crushing to obtain dry granules.
(6) And (4) adding a proper amount of lubricant into the dried granules prepared in the step (5), and sieving to prepare the whole granules.
(7) And (4) uniformly mixing the whole granules prepared in the step (6) to obtain total mixed granules.
2 method
2.1 near Infrared Spectroscopy
And (3) placing about 4g of sample in a prepared sample cup, lightly compacting, and collecting the near infrared spectrum by adopting an integrating sphere diffuse reflection mode.The scanning range is 10000-4000 cm-1Resolution of 8cm-1Gain 2, scan 64 times background air, once per hour. Each sample was scanned 3 times and the average was used for analysis.
2.2 reference value determination
The bulk density (Da) of the sample was determined by the fixed volume method using an intelligent powder property tester, the volume of the cylinder being 100ml (volume denoted V). The mass of the weighing and emptying cylinder is recorded as m1Taking a proper amount of a sample to be measured, placing the sample in a feed inlet to slowly add the sample into the measuring cylinder until the measuring cylinder is fully piled, scraping off redundant samples, and weighing the sample to be measured into m2In parallel, 3 trials were performed.
Da＝(m2-m1)/V
2.3 Spectrum pretreatment method
When the spectrum is collected, the near infrared spectrum contains redundant signals due to the influence of various factors such as environment temperature and humidity, instrument state, particle state and the like. The near infrared spectrum is subjected to proper pretreatment, so that noise can be reduced, irrelevant information can be filtered, and the robustness of the model can be improved. Such as vector normalization (normalization), Derivative (first Derivative, second Derivative), smoothing [ Savitzky-golay (sg) smoothing, Norris Derivative (ND) smoothing ], standard normal variable transformation (SNV), etc.
2.4 data processing and evaluation method
NIR spectra were pre-processed and modeled using TQ Analyst 9.0 analysis software (Thermo, USA) and samples were variable screened using Matlab 2016a (Mathwork inc., USA) software. And determining the number of main factors by using a leave-one-out cross validation method and using a cross validation Root Mean Square Error (RMSECV) as an evaluation index. And establishing a PLS quantitative model by taking the intermediate bulk density as a dependent variable and the corresponding near infrared spectrum value as an independent variable. The present embodiment evaluates the model performance with the following evaluation parameters to optimize the optimal model. The correlation coefficient (Rcal) of the correction set and the correlation coefficient (Rpre) of the verification set respectively represent the fitting degree of the correction model and the verification model, and the larger the Rcal and the Rpre are, the better the model fitting effect is; root Mean Square Errors (RMSEC) and Root Mean Square Errors (RMSEP) of prediction refer to root mean square errors between reference values and predicted values in the correction model and the verification model respectively, and the smaller the model is, the higher the prediction performance is; the correction set deviation (BIAScal) and the verification set deviation (BIASpre) respectively represent the deviation between the reference value and the predicted value in the correction model and the verification model, and the smaller the model is, the higher the prediction accuracy is. The predicted relative error (RSEP) is the relative error between the reference value and the predicted value in the model, and generally, the smaller the RSEP, the better the model prediction performance, and each evaluation parameter cannot be referred to in isolation and needs to be evaluated comprehensively.
3 results and analysis
3.1 sample partitioning
And dividing the correction set and the verification set by adopting a Random Sampling (RS) method, and ensuring that the reference value range in the verification set is contained in the correction set. Respectively randomly selecting 48 samples from the dry particle samples and the whole particle samples as a correction set, and selecting 12 samples as a verification set; the calibration set of the universal model is the sum of the two intermediate calibration sets, and the total number of the calibration sets is 96 samples, the verification set is the sum of the two intermediate verification sets, and the total number of the verification sets is 24 samples, and the division results are shown in table 15.
TABLE 15 statistical results of the reference method measurements for the calibration and validation sets of samples
3.2 selection of spectral Pre-processing method
In this embodiment, the following pretreatment methods were considered for dry granules and whole granules (120 samples in total): no pretreatment, SNV, MSC, S-G1 st, ND 1 st. Spectra were pre-treated using the above method, and the effect of different pre-treatment methods on model performance is shown in table 16. The pretreatment method was screened using RSEP as an evaluation criterion. As can be seen from Table 16, the spectral modeling performance was best using the first derivative in combination with SG smoothing pre-treatment, with an RSEP of 3.32%.
TABLE 16 Effect of different preprocessing methods on the model
3.3 feature variable screening
And the screening of the characteristic variables can eliminate irrelevant information and improve the performance of the model. The present embodiment further screens characteristic variables on the basis of the above-described screened preprocessing method.
3.3.1 screening variables based on Interval partial least squares (iPLS)
iPLS is a method in which the full spectrum is divided into several sub-intervals, and then modeling is performed in each sub-interval. In the embodiment, the full spectrum is divided into 20 subintervals, and the RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.2 screening variables based on Combined Interval partial least squares (sPLS)
The siPLS is based on iPLS, and the full spectrum is divided into a plurality of subintervals, and then the subintervals are randomly combined for modeling. In the embodiment, the full spectrum is divided into 20 subintervals, the number of combinations of the subintervals is 4, a model is established, and RMSECV is used as an evaluation index to screen the optimal modeling waveband.
3.3.3 screening for characteristic variables
The performance parameters of the model built using the 2 methods are shown in table 17. And comprehensively evaluating and screening the optimal modeling waveband by taking RMSEC, RMSECV and RMSEP as evaluation indexes. As can be seen from table 13, the original spectrum modeling is the best, probably because the full spectrum modeling contains more information and is more advantageous for building the general model.
TABLE 17 comparison of preferred Interval modeling with full Spectrum modeling
3.4 determination of the number of major factors
Improper selection of the number of primary factors results in an inadequate or excessive fit, and is generally optimized for the number of primary factors that correspond to the smallest RMSECV value. In the embodiment, a leave-one-out cross validation method is adopted, RMSECV is used as an index, the influence of the number of main factors on the model is considered, and the optimal number of main factors of the model is determined to be 10.
3.5 establishment of Universal quantitative prediction model for bulk Density of intermediates
And establishing an intermediate bulk density universal quantitative prediction model by adopting the screened pretreatment method and the spectrum band, and referring to fig. 5. The evaluation parameters of the model, Rcal is 0.905, Rpre is 0.933, and RMSEC is 0.017 g/ml; RMSECV is 0.024 g/ml; RMSEP was 0.014g/ml, RSEP was 3.23%. Rcal and Rpre are close to 1, which shows that the correlation between the reference value and the predicted value is high; the smaller RMSEC, RMSECV, RMSEP and RSEP shows that the model has higher prediction performance and can be used for the quantitative prediction of the bulk density.
3.6 model verification and evaluation
The NIR spectra of the validation set samples were introduced into a calibration model to predict the bulk density of the intermediates and compared to reference values, with the results shown in table 18. The average relative prediction error of the bulk density of the general model is 2.80 percent and less than 5 percent, which shows that the prediction accuracy of the general model is higher, and the prediction of the bulk densities of the two intermediates can be realized.
TABLE 18 near-IR reference value vs. predicted value
In the embodiment, two intermediates of dry particles and whole particles in the production process of the Yabitong capsule are taken as research objects, and a general quantitative model for predicting the bulk density of the intermediate is established by adopting a near infrared spectrum technology. Various evaluation indexes show that the universal quantitative model has high prediction accuracy, can meet actual requirements, and can realize the determination of the bulk density of two intermediates.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (10)
1. A method for establishing a universal test model of physical property indexes of an intermediate of a solid preparation is characterized by comprising the following steps:
(1) selecting a solid preparation intermediate sample, dividing the intermediate sample into a correction set and a verification set, and collecting a near infrared spectrum;
(2) determining an indicator of a physical property of the intermediate sample;
(3) pre-processing the near infrared spectrum based on the physical property indicator;
(4) after the pretreatment, characteristic variable screening is carried out;
(5) and establishing a general test model of the physical property index based on the pretreatment and the characteristic variable screening.
2. The method of claim 1, wherein the solid formulation comprises a capsule, a tablet, a granule.
3. The method of claim 1, wherein the intermediate comprises at least two of a feedstock fines, dry granules, whole granules, and total blended granules.
4. The method of claim 1, wherein the physical property indicators include particle size, angle of repose, hardness, hygroscopicity, and bulk density.
5. The method of claim 2, wherein the solid formulation is a capsule for lumbar arthromyodynia.
6. The method of claim 5, wherein the physical property index is the particle size, hygroscopicity, and bulk density of the YATONG Capsule intermediate.
7. The method of claim 1, wherein the pre-processing is selected from raw spectra or SNV pre-processing or first derivative modeling in combination with SG smoothing pre-processing.
8. The method of claim 1, wherein the feature variable screening is full band modeling.
9. A method for detecting physical property indexes of an intermediate of a solid preparation is characterized by comprising the following steps:
(1) selecting a solid preparation intermediate sample, dividing the intermediate sample into a correction set and a verification set, and collecting a near infrared spectrum;
(2) determining an indicator of a physical property of the intermediate sample;
(3) pre-processing the near infrared spectrum based on the physical property indicator;
(4) after the pretreatment, characteristic variable screening is carried out;
(5) establishing a general test model of the physical property index based on the pretreatment and the characteristic variable screening;
(6) and detecting the physical property index of the intermediate of the solid preparation to be detected by using the universal test model.
10. The method of claim 9, wherein the intermediate comprises at least two of raw material fines, dry granules, whole granules, and total blended granules; the solid preparation is a capsule for treating lumbar rheumatism;
the preprocessing is selected from raw spectrum or SNV preprocessing or modeling of first order derivation combined with SG smoothing preprocessing; and the characteristic variable screening is full-band modeling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011131359.4A CN114384040A (en) | 2020-10-21 | 2020-10-21 | Method for establishing general test model of physical property indexes of solid preparation intermediate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011131359.4A CN114384040A (en) | 2020-10-21 | 2020-10-21 | Method for establishing general test model of physical property indexes of solid preparation intermediate |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114384040A true CN114384040A (en) | 2022-04-22 |
Family
ID=81193386
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011131359.4A Pending CN114384040A (en) | 2020-10-21 | 2020-10-21 | Method for establishing general test model of physical property indexes of solid preparation intermediate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114384040A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303294A (en) * | 2008-06-20 | 2008-11-12 | 河南中医学院 | Application method of near-infrared on-line test technology in Chinese medicine Yiqing granule production |
| CN104062263A (en) * | 2014-07-11 | 2014-09-24 | 中国农业大学 | Near-infrared universal model detection method for quality indexes of fruits with similar optical and physical properties |
| CN106248620A (en) * | 2016-08-30 | 2016-12-21 | 辽宁上药好护士药业（集团）有限公司 | A kind of numbness preparation is dried the method for quick of intermediate multi objective |
| WO2018010352A1 (en) * | 2016-07-11 | 2018-01-18 | 上海创和亿电子科技发展有限公司 | Qualitative and quantitative combined method for constructing near infrared quantitative model |
| CN108195793A (en) * | 2016-12-08 | 2018-06-22 | 中国农业机械化科学研究院 | The universal model construction method of plant-derived feedstuff amino acid content |
-
2020
- 2020-10-21 CN CN202011131359.4A patent/CN114384040A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101303294A (en) * | 2008-06-20 | 2008-11-12 | 河南中医学院 | Application method of near-infrared on-line test technology in Chinese medicine Yiqing granule production |
| CN104062263A (en) * | 2014-07-11 | 2014-09-24 | 中国农业大学 | Near-infrared universal model detection method for quality indexes of fruits with similar optical and physical properties |
| WO2018010352A1 (en) * | 2016-07-11 | 2018-01-18 | 上海创和亿电子科技发展有限公司 | Qualitative and quantitative combined method for constructing near infrared quantitative model |
| CN106248620A (en) * | 2016-08-30 | 2016-12-21 | 辽宁上药好护士药业（集团）有限公司 | A kind of numbness preparation is dried the method for quick of intermediate multi objective |
| CN108195793A (en) * | 2016-12-08 | 2018-06-22 | 中国农业机械化科学研究院 | The universal model construction method of plant-derived feedstuff amino acid content |
Non-Patent Citations (1)
| Title |
|---|
| 潘晓宁: ""基于Kubelka-Munk理论的中药近红外漫反射定量研究"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》, pages 26 - 64 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yin et al. | A review of the application of near-infrared spectroscopy to rare traditional Chinese medicine | |
| Wang et al. | Evaluation of matcha tea quality index using portable NIR spectroscopy coupled with chemometric algorithms | |
| Ye et al. | Non-destructive prediction of protein content in wheat using NIRS | |
| CN1982872B (en) | Near-infrared diffuse reflection spectral method for fastly inspecting drop water content | |
| CN108519348A (en) | Licorice medicinal materials Near-Infrared Quantitative Analysis model and detection method and standard | |
| CN102288572A (en) | Method for quickly detecting content of index ingredient of traditional Chinese medicinal material by utilizing near infrared spectrum technique | |
| CN101231274B (en) | Method for rapid measuring allantoin content in yam using near infrared spectrum | |
| Wang et al. | Potential of near infrared spectroscopy and pattern recognition for rapid discrimination and quantification of Gleditsia sinensis thorn powder with adulterants | |
| Ma et al. | The rapid determination of total polyphenols content and antioxidant activity in Dendrobium officinale using near-infrared spectroscopy | |
| Woo et al. | Classification of cultivation area of ginseng by near infrared spectroscopy and ICP-AES | |
| Shikanga et al. | A novel approach in herbal quality control using hyperspectral imaging: discriminating between Sceletium tortuosum and Sceletium crassicaule | |
| Ren et al. | Hyperspectral imaging for discrimination of Keemun black tea quality categories: Multivariate calibration analysis and data fusion | |
| Ni et al. | Non-destructive detection of polysaccharides and moisture in Ganoderma lucidum using near-infrared spectroscopy and machine learning algorithm | |
| CN113655027B (en) | Method for near infrared rapid detection of tannin content in plants | |
| Zhao et al. | Geographic origin identification and rapid determination of four constituents of Gentiana rigescens by FTIR combined with chemometrics | |
| Hu et al. | Identification and quantification of adulterated Tieguanyin based on the fluorescence hyperspectral image technique | |
| CN1982873B (en) | Near-infrared diffuse reflection spectral method for fastly inspecting drop appearance quality | |
| Ni et al. | Discrimination of Radix Paeoniae varieties on the basis of their geographical origin by a novel method combining high-performance liquid chromatography and Fourier transform infrared spectroscopy measurements | |
| CN114384040A (en) | Method for establishing general test model of physical property indexes of solid preparation intermediate | |
| CN111380832A (en) | Method for constructing and detecting compound liquorice tablet effective component content determination correction model | |
| Cui et al. | Strategies for the quality control of Chrysanthemi Flos: Rapid quantification and end‐to‐end fingerprint conversion based on FT‐NIR spectroscopy | |
| CN110320286B (en) | Method for measuring content of active ingredients of Xiaochaihu granules | |
| Chen et al. | Non-destructive prediction of the hotness of fresh pepper with a single scan using portable near infrared spectroscopy and a variable selection strategy | |
| Sun et al. | Detection the quality of pumpkin seeds based on terahertz coupled with convolutional neural network | |
| CN113376116A (en) | Near-infrared online quality detection method for rehmannia |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |