Pre-formulation Studies in Drug Development

Importance of pre-formulation studies

Pre-formulation study is an important prerequisite for the development of pharmaceutical formulations. It provides an important basis for the evaluation of the drugability of compounds, the route of product administration, dosage form selection, prescription process design and quality control. It is the development of safe, effective and controllable quality drugs. A solid foundation. Including analysis and research on the physical properties, melting point, particle size, solubility, dissolution rate, crystal form, pKa, partition coefficient, surface characteristics and other inherent physical and chemical properties of the drug, the stability of the drug, the compatibility of the drug and excipients, the pharmacology, toxic side effects and irritation of the drug, as well as the laws of absorption, distribution, metabolism and excretion of the drug in vivo. These research results can help reduce potential problems in the formulation development process, ensure the safety, effectiveness and feasibility of the drug, and thereby improve the success rate of final product development.

Early-stage pre-formulation studies.

Pre-formulation studies for biologics development

Recombinant DNA technology and the concept of clusters of regularly spaced short palindromic repeats (CRISPR) associated protein 9 are flourishing in biopharmaceutical product development. These technologies, combined with artificial intelligence, can coalesce large amounts of genomic data to properly design peptides, proteins, and vaccine products. Due to the inherent instability of proteins, such as aggregation, oxidation, and deamination, the safety and efficacy of drugs may be affected. Therefore, the pre-formulation phase requires gathering all relevant information that affects the stability of the protein in solution and dry states, and determining the optimal combination either by theoretical design space or by applying surface response curve statistics. In addition, the vulnerability of peptide scaffoldings susceptible to proteolytic cleavage can be reduced by means of prodrug methods. Therefore, the peptide can be chemically altered to produce a more stable prodrug with a higher plasma concentration. Prodrugs can be produced through chemical modifications and substitution reactions such as dehydroamino acid substitution, D-amino acid substitution, thiomethylene modification, carboxyl reduction, and PEG-amino acid linking.

The design of peptide vaccines often relies on the precise identification and prediction of potential antigenic sites, including B-cell epitopes and T-cell epitopes. By using a variety of online tools and databases, researchers can predict the antigenicity, allergenicity, and toxicity of peptides and optimize their structure to enhance immune responses. For example, the IEDB Linear Epitope Prediction Tool v2 was used to predict B cell epitopes, and servers such as VaxiJen v2.0, AllerTOP v.2.0 and ToxinPred were used for screening to ensure the effectiveness and safety of peptide vaccines.

Pre-formulation studies help to select suitable adjuvants and other formulation conditions required during the manufacturing process. For example, stability analysis of a live attenuated Ty21a bacterial typhoid vaccine by spectroscopic techniques provided real-time, high-throughput information over time over a wide range of temperatures (10-85 °C) and pH (4-8). The above information is useful in the preformulation studies of other similar peptide drugs. Empirical phase maps created using data from circular dichroism and fluorescence techniques show that Ty21a cells exist in a variety of physical states, with the most stable state occurring at pH 6-7 below 30 °C. Among other potential stabilizers, 10% sucrose and 0.15 M glutamate showed the strongest protective properties, increasing the transition temperature of Ty21a cells by approximately 10°C each. The foam drying formula has also been the subject of preliminary research as a potential alternative strategy to further stabilize Ty21a cells. In addition, 10% sucrose and trehalose solutions can improve stability during processing.

Pre-formulation studies for nanoformulation development

Pre-formulation for nanoformulation development aims to design and verify the kinetic characteristics, compatibility with other substances, physical and chemical parameters and polymorphism of new drug entities to develop effective dosage forms. During the pre-formulation stage, key characteristics of the nano-drug delivery system were evaluated including shape, size, amorphous or crystal structure, and size variability, which were synthesized based on physicochemical properties. Diluents and solvents have a significant impact on the uniformity, size and shape of the nanoformulation method. Spectral analysis of active groups is crucial for the study of solubility, melting point, thermal properties and pKa behavior.

The physicochemical properties of the drug and its compatibility with excipients determine the behavior of the nanosystem selected in the final formulation design. The dissolution, polymorphic forms, pharmacokinetics, bioavailability, degradation process, adverse reaction conditions and pharmacodynamic effects of drugs are also important parts of pre-formulation studies, which rely on the physicochemical properties of the drug.

Nanoformulations are suitable for local, transdermal, injectable and oral drug delivery, and can also be used for compound development, screening, treatment, imaging and diagnosis. Consideration of factors such as polymer and adjuvant selection, key formulation analysis, preparation techniques, optimization of process variables, nanoparticle characterization, stability analysis, and improvement of encapsulation efficiency can help optimize the preparation of nanoparticles (such as lipid nanoparticles and polymer nanoparticles).

Related services at BOC Sciences

Artificial intelligence in pre-formulation studies

As one of the most important physical and chemical properties in drug discovery, lipophilicity plays an important role in regulating many key pharmacokinetic processes. Oil-water partition coefficient Oil-water P and distribution coefficient D are indicators to quantitatively characterize lipophilicity. Based on traditional lipophilicity prediction methods such as group contribution method, equation of state, and quantum chemical driving method, researchers are increasingly using AI-driven methods to associate P or D with a set of molecular descriptors, such as using charge, molecular volume and surface area, etc., for example, combining a feed-forward neural network and a KNN (k-nearest neighbor) method to predict P.

AI-assisted water solubility prediction methods predict the water solubility of drug-like molecules by learning molecular structural characteristics. For example, Coley et al. proposed graph-based convolutional neural network learning to perform convolution on the molecular map to build a specific morgan molecular fingerprint., achieved a significant lead in prediction tasks.

Machine learning models can predict the physical or chemical stability of solid dispersions. For example, Han et al. used multiple machine learning methods (including artificial neural networks, support vector machines, and random forests) to predict the physical stability of solid dispersions and found that the random forest model performed best in prediction with an accuracy rate of up to 82%. This method can effectively predict stability under different formulations and optimize formulation design.

Related services at BOC Sciences

Unbalanced drug absorption, distribution, metabolism, excretion (ADME) and toxicity characteristics are common reasons why candidate drugs cannot be successfully advanced in the later stage, and may even lead to the withdrawal of approved drugs. Shin et al. proposed a DNN model that avoids complex feature engineering to predict the permeability of compounds with multiple structures on Caco-2 cells. Sun et al. used 6 machine learning methods and 26 physical, chemical and structural descriptors to build a prediction model. The prediction mean absolute error (MAE) range is 0.126 to 0.178, which has excellent performance. In terms of metabolic site prediction, there are studies that model the formation of reactive metabolites based on deep learning networks and estimate the metabolism of various chemical substances.