Data-driven Intelligent Design Of Traditional Chinese Medicine High Shear Wet Granulation Process

Wet granulation is the process forming particles with a certain structure and morphology,and is widely used in the production process of oral solid preparations of traditional Chinese medicine.High shear wet granulation(HSWG)is a prevailing granulation technique.After the drug and excipients are mixed uniformly,the powder bed and the binder interact under the action of shearing force,and the particles are wetted.It products particles in three simultaneous stages:wetting and nucleation.growth and consolidation,attriction and breakage.The HSWG process is affected by many factors,such as the attributes of the raw material,the physical properties of the binder liquid,the amount and the addition method of binder,impeller and chopper speed,and the time of wet mixing,etc.At present,there is no recognized HSWG process mechanism model.In order to realize the HSWG process design,this dissertation adopts a combination of data-driven approaches and process modeling methods under the guidance of the Quality by Design(QbD)discipline,and establishes a method for predicting the particle size of HSWG made from literature data and experimental data.Partial least squares(PLS)model was employed to explore the manufacturability classification rules of Chinese herbal extract powder and common used medicinal excipients in the process.Finally,based on the process model and simulation algorithm,the formulation and process parameter design method in line with the characteristics of traditional Chinese medicine was established and verified by experiments.The specific research content includes the following three parts:1.Research on high-shear wet granulation based on literature dataWe have retrieved total 249 published articles,collected 143 sets of data form the filtered 10 artcles,including 17 formulations that was consisted of total 22 materials(19 excipients and 3 active pharmaceutical ingredients).The HSWG experiments use water as a binder liquid.The granulation scale is distributed from 0.25 L to 600 L.The materials collected from literatures and iTCM database are matched according to certain rules.At the same time,a dimensionless parameter,Froude number,is used to integrate different granulation process parameters to eliminate the influence of granulation scale.By fusing different sources of data,the PLS regression method was built to establish an integrated function of physical properties of formulation,process parameters,powder-binder interaction paraters and particle quality.The determination coefficient R2 of the pure literature data model is 0.743,the prediction ability Q2 is 0.718,and the root mean square error of prediction(RMSEP)is 152.6,and 36.2%of the samples have prediction errors less than 100 ?m.The augmentation experiments are designed further to improve the prediction performance of the model.The determination coefficient R2 of extended model is 0.710,the prediction performance Q2 is 0.680,and the RMSEP reduced to 115.7,and prediction errors of 43%of the samples were less than 100 ?m.The raw material collected in the literature composed a material library,and exploratory principal component analysis can be used to show its diversity.By comparing the PLS regression model established purely from literature data with the model after added experiment data,it was found that adding a small amount of experiment data into the model can significantly reduce its prediction error,especially in the case of a lower binder content(liquid-solid ratio less than 0.6).The data fusion method used in this part can promote process of formulation and process optimization in HSWG,and can save the time and cost of the experiment.2.Study on manufacturability classification of materials for Traditional Chinese Medicine in HS WGThis part built a small material library composed of 20 kinds of materials,including 8 kinds of Chinese medicine extract powder and 12 kinds of excipients.Each material is described by 22 kinds of physical parameters.Binders with different hydrodynamic properties were considered,and the parameters of density,viscosity and surface tension were characterized.Combining the empirical constraints and D-optimal experimental design,56 granulation experiments were arranged,and the granulation experiments were carried out on a 2 L scale.The PLS algorithm was used to establish a process model for predicting the accumulative median particle size(G50)based on the properties of materials and binders.contact parameters and process parameters.The surface tension and density of the binder liduid,as well as the process dimensionless parameter maximum pore saturation were identified as the key variables.In the latent space of the optimized HSWG process model,the materials are divided into three categories according to the material distribution characteristics,namely,Chinese medicine extract powder,diluent and disintegrant.The granulation of traditional Chinese medicine requires comparatively lower viscosity and small dosage of binder,and the resultant particle size is small,such as 75%,85%and 95%ethanol;diluent powders have a larger physical space and can be made into particles of different particle sizes by adjusting different kinds of binder;the disintegrant can easily be made into large-size particles under the condition of an aqueous binder.The combination of the material library and the multivariate modeling method not only expands the physical properties of the input materials of the model,but also facilitates the generalization of knowledge of the wet granulation process of Chinese medicine,and provides a basis for the formulation and process design based on the properties of the materials.3.Model-driven methodology for HSWG process design of traditional Chinese medicineIn order to deeply explore the behavior of different types of materials and mixing systems in the HSWG,and under the proper constraints(G50 in the range of 100-300 ?m),the design space is delineated to achieve the goal of binder and granulation condition selection.The purpose of this part is to generate a total of 2931 data through simulation,including 1080 single-materials,891 binary mixed formulations,and 960 ternary mixed formulations.4 kinds of Chinese medicine extract powder(Mentha haplocalyx Briq,Pogostemon cablin,Lindera aggregata and Cinnanmomam cassia Presl)and 2 kinds of commonly used excipients(MCC PH101 and lactose Granulac 200)are used as model drugs.The impeller and chopper speed.and wet mixing time were simulated,then the maximum pore saturation was calculated,and the contact angles between the mixed powder and the binder were matched respectively.The simulation results showed that when the concentration of Chinese medicine is greater than 80%,75%ethanol or water with relatively larger surface tension and viscosity should be used as the binder(water and 75%and 95%ethanol)to gain the ideal granules under the reasonable particle size constraints.With the increase of excipients,the particle size increases.And the bonder should be replaced with 95%ethanol.It is recommended to use 95%ethanol as the binder for the formations of the ternary mixed system of Chinese medicine extract powder,MCC and lactose.The higher viscosity,such as water and 4%HPMC,tend to result in larger particles.In addition,this part also organized a total of 12 verification experiments,including 4 single material samples,5 binary mixed samples and 3 ternary mixed samples.As the verification results show that the model has a good prediction performance for Chinese medicine extract powder with an average error of 20.7 ?m,while the pure excipient and the mixed system of Chinese medicine and excipients have a larger prediction error of 221.9 ?m due to the different granulation mechanisms.In conclusion,under the guidance of quality by design(QbD),this work focused on predicting the critical quality attributes of particles,collected a large number of literature data and identified key variables in combination with experiment design,and established a model integrating material properties,binder properties,process parameters and contact parameters to predict critical quality attribute of granule.Through this model,multi-formulation design,automatic selection of binder liqud and intelligent design of process parameters collaboratively are realized.At the same time,it innovatively proposes the classification of material manufacturability of HSWG,distinguishing the wet granulation behavior and conditions of different materials,by which the PLS model strengthens the understanding of the process.Finally,the Monte Carlo simulation technology is used to explore the design space of a single material and a multi-component system,and the robustness of the methodology is verified through experiments.