Poster X3 - Machine Learning-guided excipient derivatization for self-assembling drug nanoparticles.

Introduction: Enhancing the bioavailability and efficacy of small molecular therapeutics requires advanced drug delivery solutions. While solid drug nanoparticles stabilized by lipids and polymers offer a solution, their limited drug loading poses an issue. Recent advancements in combining cancer drugs with specific small molecular dyes have shown promise in creating stable nanoparticles with high drug loading. However, only a limited number of these small molecule excipients are known, they are unable to encapsulate all drugs with sufficient stability and can have additional safety concerns including immunogenicity and carcinogenicity. Therefore, we require innovative approaches such as machine learning (ML) techniques to create new excipients for augmenting our material armamentarium to improve nanoparticle stability and effectiveness. Here, we let machine learning models guide excipient derivatization via chemical synthesis to create novel excipients with improved ability to encapsulate more drugs at greater stability. Taken together, machine learning-guided molecular derivatization holds great potential for the creation of novel drug delivery solutions as well as enhance other optimization processes in chemical and biological engineering.

Materials and

Methods: In this study, random forest ML algorithms were employed to analyse and derive nanoparticle forming excipients using data from Reker, D., et al. Nat. Nanotechnol. 16, 725–733 (2021). The dataset encompassed various excipients and their properties, alongside nanoparticle formation outcomes. We then created a virtual library of excipient molecules by systematically derivatizing their molecular structures in silico. This virtual library was then screened with help of the ML model to prioritize novel excipients with high likelihood to serve as nanoparticle-forming stabilizers for synthesis and testing. Briefly, excipients were synthesized using batch chemistry and purity was assessed using NMR. Successfully synthesized excipients were then experimentally evaluated for their efficacy in forming stable nanoparticles with self-assembling drugs using dynamic light scattering (DLS).

Results, Conclusions, and Discussions: Our ML-assisted approach successfully identified 23 novel excipient derivatives with enhanced nanoparticle-forming capabilities from a virtual library of 475 novel excipient designs.. These derivatives exhibited favorable molecular properties conducive to self-assembly with drugs, resulting in stable nanoparticle formations. Dynamic light scattering (DLS) measurements confirmed the predicted nanoparticle stability and efficiency of the designed excipient derivatives. Further, the reference identity of the compounds used in this study cannot be disclosed at this time due to pending patent filing but will be disclosed at the time of the presentation. Together, these findings hold significant promise for advancing drug delivery systems by enabling the design of excipients optimized for nanoparticle formation through machine learning, thereby rationally creating novel materials to improve nanoparticle formation, stability, drug loading capacities, and ultimately therapeutic outcomes.

Acknowledgements (Optional):