Implementation of new machine learning algorithms for the optimisation of drug formulations

Project Information

MACHINE-DRUG

Grant agreement ID: 101123353

DOI

10.3030/101123353

Project closed

EC signature date 2 June 2023

Start date 1 October 2023

End date 31 March 2025

Funded under

• European Research Council (ERC)

Total cost

No data

EU contribution

€ 150 000,00

Coordinated by

UNIVERSITE DU LUXEMBOURG
Luxembourg

Periodic Reporting for period 1 - MACHINE-DRUG (Implementation of new machine learning algorithms for the optimisation of drug formulations)

Reporting period: 2023-10-01 to 2025-03-31

Summary of the context and overall objectives of the project

Correctly developing and predicting crystalline forms with specific physico-chemical properties is essential for the pharmaceutical industry. Polymorphism, where an active pharmaceutical ingredient (API) can exist in multiple crystalline forms, poses a significant challenge, as different forms can have varying solubility, release rates, and therapeutic efficacy. Unforeseen polymorphs, especially those that emerge after drug approval, can have serious economic and health consequences.
Currently, crystal polymorphs are assessed through computational molecular crystal structure predictions (CSP), but these tools face challenges due to the complexity of crystallographic space and the need for highly accurate calculations. The goal of this project is to accelerate and improve the crystal structure prediction process using novel machine learning methods, making it more efficient and accessible for the pharmaceutical industry.
By developing a platform that integrates machine learning with vibrational spectra simulations for polymorph identification, this project accelerates the process by a factor of 100, significantly reducing the time and cost involved. This advancement enables the pharmaceutical industry to more accurately predict and manage crystalline forms, ultimately enhancing drug formulation and minimizing associated risks.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The project successfully focused on the development and implementation of a cutting-edge platform designed to accelerate the identification of polymorphs in molecular crystals. The following key activities were performed:
• Development of an Automated Platform (THeSeuSS) for Vibrational Spectra Simulation:
The core activity involved the creation of an automated platform that simulates vibrational spectra, which play a crucial role in the identification of different polymorphs. This platform has the potential to be widely adopted in both academic and industrial settings for improving drug formulation and understanding material properties.
• Integration of Machine Learning-Based Force Fields:
In line with the project objectives, machine learning force fields were incorporated into the platform. These force fields, developed by the PI’s research group, were designed to also improve the accuracy and efficiency of the polymorph prediction process.
• Optimization for Scalability and Efficiency:
The platform was designed to handle large systems efficiently, reducing the time and cost involved in identifying polymorphs. The integration of machine learning techniques allowed for an accelerated process, in both time and cost compared to conventional methods.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

The project has delivered a highly functional and scalable platform for the simulation of vibrational spectra, designed to assist in the identification of polymorphs in molecular crystals. By integrating machine learning techniques, the platform accelerates the identification process, reducing both time and cost.


Key Results:
• Automated Vibrational Spectra Simulation Platform: The platform is capable of simulating and analyzing vibrational spectra to identify polymorphs, a crucial task in materials and pharmaceutical research. This tool opens new avenues for more efficient and reliable polymorph screening.
• Integration of Machine Learning Models: The integration of machine learning force fields has significantly enhanced the accuracy and efficiency of the polymorph identification process.
• Open-Source Access: The platform is provided as open-source software, which promotes wide adoption across the research and pharmaceutical communities. This open-source approach fosters collaboration and further development by the global scientific community, helping to advance the field.
Potential Impacts:
• Improved Pharmaceutical Development: By streamlining the process of identifying polymorphs, the platform will have an impact on drug formulation. It will help pharmaceutical companies better manage the risks associated with polymorphism, leading to more stable and effective drug formulations.
• Cost Reduction and Efficiency: The reduction in time and cost for polymorph identification will enable faster time-to-market for new drugs. This not only benefits the pharmaceutical industry but also enhances global healthcare access by reducing drug development timelines and costs.
• Advancements in Material Science: Beyond pharmaceuticals, the platform has applications in materials science, where the prediction and identification of crystal forms can impact industries such as battery technology, manufacturing, and chemicals.

Key Needs for Further Uptake and Success
To ensure that the platform reaches its full potential and has a lasting impact, several key actions and areas of support are needed:
1. Further Research and Development:
While the platform has shown significant promise, continued research is required to refine the machine learning models, improve prediction accuracy, and expand the platform’s capabilities to include a wider range of methods. Research into enhancing the platform's usability and scaling for industrial use is also crucial.
2. Commercialization and IPR Support:
Although the platform is open-source, there is potential for commercialization in the form of advanced, enterprise-level solutions. To facilitate this transition, intellectual property (IPR) support and licensing strategies will be needed to protect and monetize the platform, while still allowing for broad community engagement.
3. Supportive Regulatory and Standardization Framework:
Establishing a supportive regulatory framework will be critical to the widespread adoption of computational tools in pharmaceutical and materials development. Collaborating with regulatory bodies to ensure the platform meets industry standards and regulatory requirements will be key to its success.
Overview of the Results
The project has delivered a robust and scalable platform for simulating vibrational spectra and identifying polymorphs, integrating machine learning force fields to reduce the time and cost of polymorph identification. The open-source nature of the platform ensures its broad accessibility, allowing for further advancements and use in pharmaceutical development and materials science. By addressing the significant challenges of polymorphism in drug development, this project contributes to improving drug formulation, accelerating time-to-market, and reducing associated risks.

machine-drug-image.jpg