Research On Quantitative Model For Fundamental Data And Structure-activity Relationships Of Premium Chemicals

The fundamental data in chemical engineering,including physical properties（especially phase equilibrium）and kinetic data,are important foundations for process simulation.However,with the extension and intersection of chemical engineering to multiple disciplines,physical property research has transcended the typical organic small molecules involved in the traditional chemical industry,and non-traditional physical property research is oriented towards premium functional chemicals with complex nano structures.This has led to challenges for the prediction methods of traditional fundamental data that combine small amounts of experimental data with thermodynamic models.The main reason for this is that the complex nano structure leads to"difficult and inaccurate measurement"of the fundamental data.However,the research and development of premium functional chemicals often achieve excellent performance through many experiments to control their nano structure and interface properties,and a large amount of fundamental data in the research and development process is kept confidential.These contradictions lead to a serious lack of fundamental data in the development of premium functional chemicals,and the structure-activity relationship is unclear.Therefore,the key to innovative development and intelligent manufacturing of premium functional chemicals is to develop new paradigms to obtain fundamental data under the influence of nano structures and quantify the relationship between nano structures and performance.This paper focuses on pharmaceutical formulations and uses theoretical calculations,chemical engineering thermodynamics,and machine learning to model the thermodynamic behavior of drugs and polymers,determine the interaction mechanism between drugs and polymers,and reveal the key interactions that affect physical properties.To study the thermodynamic properties of pharmaceuticals under the influence of typical nano structures such as micelles and nanocrystals,we constructed a complex interface thermodynamic model that quantitatively correlates Gibbs energy with structural and molecular parameters of surfactants/stabilizers.Based on the key interactions that affect physical properties,we purposefully screened a small number of key intermolecular interaction descriptors based on molecular thermodynamics,constructed interpretable machine learning models that have good generalization performance and high accuracy in predicting fundamental data.To study the thermodynamic properties of pharmaceuticals under the influence of typical nano structures such as micelles and nanocrystals,we constructed a thermodynamic model of complex interfacial systems in which Gibbs energy can quantitatively Gibbs energy can quantitatively correlate with the structural and molecular parameters of surfactants/stabilizers.Further,the drug release performance under the influence of nano structure was studied,and a quantitative model for predicting drug release rate was established coupled with internal diffusion and external diffusion.The main research contents are as follows:The phase behavior and interaction mechanism of model drugs probenecid（PBD）and flurbiprofen（FBU）with polyethylene glycol（PEG 6000）,polyvinylpyrrolidone（PVP K30）,hydroxypropyl methyl cellulose（HPMC E3）,and copolyretinone（PVPVA 46）were studied using differential scanning calorimetry（DSC）and density functional theory（DFT）.It was found that the stronger the hydrogen bond（HB）interaction,the greater the solubility of PBD and FBU in the polymer,and therefore both PBD and FBU have the highest solubility in PVP K30.The thermodynamic model of complex interface systems in which Gibbs energy can quantitatively correlate with the structural and molecular parameters of surfactants is constructed.To study the thermodynamic properties of drugs under the influence of micelle nanostructure,biorelevant media was selected as the research system.A pH dependent solubility model was firstly proposed.Compared to the Henderson Hasselbalch equation without considering buffer composition,the average error of modeling PBD solubility in blank fasted state simulated intestinal fluid(Fa SSIF_Blank)and blank fed state simulated intestinal fluid(Fe SSIF_Blank)was reduced from 18.94%and 14.92%to 0.91%and 1.92%,respectively.The average error of modeling for the solubility of FBU in Fa SSIF_Blank and Fe SSIF_Blank decreased from 50.96%and 37.32%to 0.59%and 0.59%,respectively.A micellar solubilization model was proposed to accurately model the drug solubility in biorelevant media,with a maximum average error of 2.45%.The thermodynamic model of complex interface systems in which Gibbs energy can quantitatively correlate with the structural and molecular parameters of stabilizers is constructed.The quantitative relationship between the solubility of nanoparticles coated by stabilizers and the particle size,interface energy was determined by solid-liquid phase equilibrium analysis.According to the quantitative relationship and the solubility of nano metals,the expression of interface energy based on the molecular parameters of stabilizers and nanoparticles was determined.It was found that when the radius of the nano metal was less than40 nm,the ratio of interface energy relative to the Gibbs energy enhancement is more than 50%and 30%for Ag and Cu coated by stabilizer,respectively,and the interface energy played an important role.Compared to the Ostwald Freundlich equation（with a maximum relative deviation of 12.98%and a minimum relative deviation of 5.45%）,the interfacial thermodynamic model is more accurate in predicting the solubility of nano drugs,with a maximum and minimum relative deviation of only 2.70%and 0.87%,respectively.According to the interaction mechanism between drugs and solvents/polymers,a method of directed screening molecular descriptions based on interaction mechanisms was proposed.Compared with the usual>100 descriptors,only 16 and 11 descriptors were needed to train machine learning models that have good generalization performance and can predict fundamental data.Molecular thermodynamics used for data augmentation has improved the performance of machine learning models,increasing the R² of drug solubility training and testing sets from 0.92 and 0.81 to 0.99 and 0.98,respectively.The importance of drug solvent binary interaction descriptors in drug/solvent systems is positively correlated with the polarity of the solvent.In drug/polymer systems,hydrogen bond interaction descriptors are more important than molecular structure and other intermolecular interaction descriptorsTo study the drug release performance of sustained-release solid dispersion under the influence of nano structure,a quantitative model for predicting drug release rate was established coupled with internal diffusion and external diffusion.A series of FBU sustained-release solid dispersions were prepared by solvent evaporation.X-ray diffraction（XRD）and differential scanning calorimetry（DSC）confirmed that FBU existed in the carrier in an amorphous form.The dissolution kinetics of FBU sustained-release solid dispersions in a pH 6.8 buffer solution was measured to study the effects of Eudragit RL and Eudragit RS ratios,drug loading,additive types,and additive ratios on the dissolution kinetics.The FBU release process is accurately modeled through the coupling of internal and external diffusion,with a maximum average error of 11.8%.It is revealed that the resistance ratio of internal and external diffusion is much greater than 10⁴,and the FBU release rate control step is internal diffusion.The internal and external diffusion dissolution kinetics model quantitatively correlates the relationship between nano structure of pharmaceutical formulations and drug release rate,and predicts the dissolution kinetics of FBU sustained-release solid dispersions with different formulations and particle sizes.