FormulationMM: A universal computer-driven drug formulation platform

Abstract

Classical computer-aided drug design answers “Will this ligand bind?”, whereas computer-driven drug formulation answers the downstream but equally critical question “How can this drug be physically made and delivered?”, thereby bridging the long-standing gap between molecular modeling and drug-formulation science through physics-based and fully automated multiscale simulation. To build this bridge, we introduce FormulationMM, an innovative platform utilizing real-world driven molecular modeling to explore drug formulation mechanisms. FormulationMM features a pharmaceutical formulation algorithm, a comprehensive excipient database, and robust modeling protocols, ensuring a streamlined workflow for the generation, simulation, and analysis of drug formulation. It automatically generates forcefield parameters for drug molecules and excipients, supporting six formulation types: cyclodextrin-drug inclusion, micelles, liposomes, solid dispersions, self-assembling drug nanoparticles, and transmembrane drug delivery systems. Our results closely match experimental findings and demonstrate high predictive accuracy and reliability. FormulationMM, accessible through a continuously updated website (https://formulationmm.computpharm.org), offers a practical platform to support drug formulation research and development, with the potential to advance the growing field of computer-driven drug formulation.

Introduction

The contemporary field of drug development focuses not only on the efficacy of drugs but also on optimizing their delivery to enhance therapeutic results while reducing side effects and toxicity. This requires the development of innovative drug delivery systems (DDSs) and formulation methods [[1], [2], [3]]. DDSs are designed to enhance the therapeutic efficacy and safety of drugs by controlling their release, targeting specific tissues or cells, and minimizing side effects [1,2]. Current DDSs are highly complex and varied, encompassing cyclodextrin-drug complexes, solid dispersions, micelles, liposomes, drug nanoparticles, and transmembrane drug delivery systems, among others. DDSs experiments often entail numerous blind tests due to unclear mechanisms at the molecular level. Consequently, the design and validation of DDSs primarily rely on empirical methods, which involve extensive experimentation and trial-and-error processes [4]. This traditional approach is often time-consuming, costly, and inefficient, highlighting the need for more precise and predictive tools.

The complexity and diversity of DDSs present significant challenges to both traditional modeling scientists and pharmaceutical scientists. Molecular modeling with the physics-driven principle has emerged as a valuable method in computer-aided drug design (CADD) [[5], [6], [7], [8]] and computer-driven drug formulation (CDDF) [4]. However, significant differences exist between CADD and computer-driven drug formulation CDDF [4], as summarized in Table 1. In this case, computer-driven drug formulation offers a theoretical and computational approach to understanding and predicting the behavior of drugs within various drug delivery systems at the molecular level, based on the principles of classical mechanics and statistical mechanics. On the one hand, due to the lack of molecular modeling tools specific to DDSs, pharmaceutical scientists currently struggle to deeply understand the underlying molecular mechanisms of these systems. For example, molecular modeling demand a solid grasp of molecular mechanics principles, proficient programming abilities, and access to advanced computing facilities. The synergy of theoretical understanding, computational techniques, and powerful computing infrastructures is crucial for conducting in-depth simulations that can forecast behaviors and interactions of DDSs with precision. In addition, generating precise molecular force field parameters is essential for authentic molecular modeling. This task requires sophisticated knowledge of quantum chemistry, particularly as the complexity increases with the size of the molecules involved, from small-molecule drugs to larger excipient substances. This is a complex and challenging process that includes calculating atomic charges and generating bond, angle, dihedral, and Lennard-Jones potentials. Researchers need to carefully choose the level of computational methods to strike a balance between the computational cost and the accuracy of the simulation.

This disconnect highlights a critical knowledge and tooling gap between pharmaceutical scientists and molecular modelers. On the other hand, researchers in the molecular modeling field are proficient in theoretical computations related to molecular simulations. However, without pharmaceutical experience, it is challenging for them to pinpoint the research focus in DDSs simulations and effectively apply their skills to DDSs projects.

Simulation methodologies often vary with different DDSs, making the selection of appropriate methods critical and largely dependent on the nature and size of the DDSs. Different DDSs, such as cyclodextrins, nanoparticles, solid dispersions, transmembrane systems, and larger entities like micelles and liposomes, necessitate distinct simulation techniques. Moreover, researchers must have a deep understanding of the dynamics and physical characteristics of each DDSs to choose the appropriate force fields and simulation tactics. This includes considering the manufacturing method of DDSs, their inclusion complexes, and delivery mechanisms, as well as other important phenomena. This customized strategy, specific to the relevant formulation, is instrumental in uncovering key elements that affect the success of drug delivery, thereby helping to focus research efforts on the most impactful outcomes.

Since 2011, we have successfully developed the molecular modeling method to decipher the mechanism of various types of DDSs at the atomistic and molecular level, including solid dispersion [9], cyclodextrin-complex [10], liposome [11], micelle [12], and self-assembly drug nanoparticle [13]. The book <Computational Pharmaceutics – applications of molecular modeling in drug delivery> in 2015 is the first monograph to illustrate and discuss the application of computational methods in pharmaceutics [14]. Then we continued to publish < Exploring Computational Pharmaceutics − AI and Modeling in Pharma 4.0 > in 2024 to provide an extensive and up-to-date overview of the theory and application of computational pharmaceutics in the drug development process [15]. However, despite this substantial body of work, our previous research efforts were largely fragmented, system-specific, and manually executed, lacking the integration, automation, and scalability required for high-throughput formulation design. There was no unified pipeline to streamline the modeling of different DDS types, nor a platform capable of delivering consistent and reproducible results across varying systems and use cases.

FormulationMM was developed to bridge this gap. This platform unifies the fragmented modeling approaches we previously used and transforms them into a streamlined, modular workflow that can automatically support major types of formulation strategies. In addition to integrating our accumulated knowledge, FormulationMM provides reproducible and scalable results that help reduce user dependence on expert-level modeling skills. We anticipate that it will contribute to the advancement of computational pharmaceutics by improving accessibility, accelerating design cycles, and enhancing the reliability of predictive modeling for drug delivery research.