Using Advanced Algorithms to Solve Formulation Challenges

AI-Enhanced Medicine Pill Concept for Future Healthcare. Pill symbolizing integration of artificial intelligence in the development of future healthcare and advanced treatments | Image Credit: ©PSCL RDL- stock.adobe.com

Artificial intelligence (AI) and machine learning (ML) in the biopharmaceutical industry have been widely applied across drug discovery and clinical development efforts. These technologies also have potential to impact many other aspects of drug development and manufacturing. Their use in formulation development has been growing in recent years due to the potential for AI/ML to streamline and accelerate this process, particularly for challenging drug substances such as those that suffer from poor solubility and bioavailability (1).

The traditional approach to formulation development involves extensive, time-consuming, and costly physical experimentation and screening, both in vitro and in vivo, to identify optimal formulations. Leveraging AI/ML algorithms allows for identification of formulations with the greatest likelihood of success, reducing the time and resources needed for physical evaluation. These technologies can also uncover new materials with attractive excipient properties and new formulation approaches not previously considered (2).

One example of an AI-based prediction tool designed specifically to aid formulation design is the freely available web-based platform, FormulationAI (3).

Formulation development plays a key role

Formulations play a critical role in stabilizing APIs, enabling patient administration, and potentially mitigating toxic side effects while enhancing drug efficacy. “Regrettably,” observes Christine Allen, a professor within the Leslie Dan Faculty of Pharmacy at the University of Toronto and CEO of Intrepid Labs, a biotech start-up that utilizes AI/ML and robotics to accelerate formulation development, “the significant potential of formulations is often underestimated and, consequently, given lower priority in drug development, despite their ability to directly improve safety and efficacy.” She notes that approximately 80% of drugs fail in clinical development due to safety concerns and lack of efficacy, areas where better-formulated drugs might succeed.

Poor solubility/bioavailability present many development challenges

Between 70–90% of new chemical entities (NCEs) have solubility issues, according to Allen. A formulation that excels in enhancing solubility and bioavailability but falls short in safety, efficacy, or patient acceptability does not fulfill its therapeutic potential, adds Sanjay Konagurthu, senior director of science and innovation within Thermo Fisher Scientific’s Pharma Services business.

“One of the greatest challenges, therefore, is for developers to meticulously balance these considerations, creating formulations that not only improve drug delivery but also meet the broader criteria necessary for successful patient outcomes,” Konagurthu contends.

Speed is another ongoing challenge in the journey from molecule to medicine. “While maintaining safety and efficacy, scientists are always looking for a faster way to overcome hurdles like poor solubility and bioavailability and the trial-and-error formulation development process to get therapies into the hands of patients as soon as possible,” explains Konagurthu.

Difficulties arise, Allen adds, because current formulation methods are often slow and resource-intensive, which can hinder the rapid development and optimization of drug formulations that are crucial for overcoming solubility challenges.

In fact, Konagurthu comments, because poor API solubility typically corresponds with lower bioavailability and thus impacts the efficacy of the drug product, it can create a negative ripple effect from discovery to clinical trials. He points to several factors, from molecular properties to the characteristics of the excipients and formulations, as determining the solubility and bioavailability of an API. “For this reason, identification of the most effective combinations of drug and excipients is a complex and resource intensive process that requires rigorous experimentation, data collection and analysis,” he says.

Historically, scientists have looked to resolve poor API solubility through trial-and-error testing, which often results in longer development timelines. “With pharmaceutical companies focused on getting life-saving therapies to patients as quickly as possible, there is an urgent need for streamlined workflows that allow scientists to efficiently find the best way to improve a drug’s bioavailability,” Konagurthu concludes.

Predictive capabilities offer many benefits

Advanced predictive algorithms can help overcome this important challenge. “AI/ML offers low-cost predictions of properties such as solubility, dissolution rate, and stability,” Allen says. “Importantly, advanced AI approaches, such as those employed at Intrepid Labs, no longer rely on data-hungry ML models. In addition, combining technologies such as autonomous preparation and real-time sensing further enhances data-driven processes,” she adds.

In addition to predicting the most effective combinations of solubility enhancement technologies and excipients for drug development, Konagurthu believes use of AI/ML in this type of application can help scientists better understand the complex behavior of molecules, improve the speed of API discovery, and increase the accuracy of formulation development. “Ultimately, because these new technologies are able to generate high-quality data and insights for use throughout the drug discovery and development process, using them can have a significantly positive impact on budgets, timelines, and resources,” he states.

Varying data types are useful

Another advantage of using AI/ML in formulation development is the ability to leverage wide-ranging datasets depending on the prediction target, route of administration, and so on. To address poor solubility, for instance, Allen notes that the apparent solubility or dissolution rate for amorphous solid dispersions (ASDs) can be predicted using ML models trained on datasets containing the composition of the ASD formulations, molecular descriptors, or other representations of the API(s) and polymers, and additional information on the method used to make the formulations, such as the processing parameters and attributes of the solution used for solubility or dissolution testing.

In many cases, computational chemistry methods including quantum mechanics (QM) and molecular dynamics (MD), quantitative structure-activity relationship models, and aspects of ADMET (absorption, distribution, metabolism, excretion, and toxicity) analysis are being applied to create customized formulation predictions as well, according to Konagurthu. To address poor solubility specifically, he adds that many solutions use proprietary algorithms to create predictive models based on specific properties of the substances. These tools are then used to generate predictive models for solubility and bioavailability enhancement; accelerated stability models for shelf-life and packaging determination; materials science, compaction simulation, and process models; and ADME/PK (pharmacokinetic) models to predict the effects of API physicochemical properties and pharmacokinetics.

For example, Konagurthu observes that techniques like QM/MD offer modeling capabilities that provide deep insights into physicochemical properties and molecular-level interactions, enabling structural mechanism exploration, free-energy evaluation, and spectroscopic characterization. “Using these insights in combination with AI/ML approaches has proven to be accurate and more efficient than trial-and-error experimentation,” he adds.

Application in formulation development is increasing

Over the past two decades, there has been a substantial increase in literature focused on the role of computational chemistry in the pharmaceutical industry, which highlights the capabilities of AI/ML in formulation development, according to Konagurthu. “The literature suggests an uptick in adoption and gives the scientific community proof points of how these new technologies address long-standing industry challenges of solving poor API solubility and bioavailability,” he notes.

One way to infer the level of adoption of AI/ML, Konagurthu says, is to look at the increase in regulatory submissions that involve these methods. Since 2021, FDA has received more than 100 submissions for drug and biologic applications using AI/ML components, spanning the drug development landscape from molecule to medicine (4). This surge in interest prompted FDA to create specialized regulatory groups that serve as both resources and facilitators of innovation for the pharmaceutical industry, he notes.

Specifically considering the development of formulations for poorly soluble/bioavailable APIs, Allen observes that the use of AI/ML is less advanced than it is for other areas of drug development, but it is growing.

Some hurdles still to clear

To further the use of AI/ML solutions for formulation development, there are many hurdles to clear. As is the case with any data-driven application, large quantities of relevant, high-quality data are needed to create and train AI/ML models for formulation development. “Alternative ML approaches that require less data should also be evaluated. To see real success in this area, however, there is a need to embrace integration of AI/ML platforms within existing pharmaceutical frameworks,” comments Allen.

Other challenges noted by Konagurthu include model interpretability, intellectual-property concerns, cost of implementation, workforce training, and the need for regulators to develop and establish standards that not only help to maintain the integrity of these computational methods, but also help governing bodies evaluate submissions that use AI/ML in drug development.

These hurdles can be overcome, according to Allen, by building more interdisciplinary teams combining AI/ML expertise with pharmaceutical knowledge and increasing investment in AI/ML capabilities and training within the pharmaceutical sector​​. “Concerted efforts in data management, regulatory engagement, technology integration, and training will ensure that the benefits of AI and ML can be fully realized in pharmaceutical formulation development,” Konagurthu agrees.

Partnerships can be valuable

Drug developers can also better leverage AI/ML platforms by identifying and working with a strategic contract development and manufacturing organization (CDMO) with deep experience and a strong track record of managing challenges posed by new technology adoption, according to Konagurthu. “For scientists who are looking to address poor API solubility and bioavailability, it can be extremely beneficial to collaborate with a CDMO specializing in computational drug development. CDMO partners are committed to staying at the forefront of pharmaceutical innovation and are adopting new technologies focused on speed, scalability, and innovation as a part of their end-to-end offerings,” he says.

Such partnerships can be highly effective in supporting pharmaceutical companies across the clinical supply chain, helping to ensure the faster delivery of high-quality therapeutics to patients, Konagurthu contends. He also emphasizes that widespread collaboration between pharmaceutical companies, CDMOs, and regulators on next-generation technologies could have many positive benefits for the pharmaceutical industry and the patients it serves.

AI/ML poised to have a significant impact

Both Allen and Konagurthu expect to see valuable short- and long-term benefits realized due to the use of AI/ML technologies to enable formulation development for APIs with poor solubility and bioavailability. For instance, working with colleagues at the University of Toronto, Allen has demonstrated that ML algorithms can be used to predict experimental drug release from long-acting injectable formulations and help guide their design (5).

“Using AI/ML technologies to address solubility and bioavailability challenges is a significant shift for the pharmaceutical industry. These technologies offer a more streamlined, accurate, and cost-effective path forward for drug development in general. For poorly soluble APIs in particular, they are enabling optimization of the selection of solubility enhancement technologies and excipients and paving the way for a more efficient and sustainable process,” states Konagurthu.

In the short term, Allen expects the enhanced predictive capabilities afforded by AI/ML platforms will lead to more efficient formulation processes. Longer-term, she anticipates that as drug formulation datasets reach critical mass and formulation scientists become more well-versed in AI/ML, ongoing advances in these technologies will continue to unlock new potential in formulation science, requiring continuous adaptation and learning within the industry. “There is tremendous potential for AI/ML to enable the accelerated discovery of new drug formulations and innovative drug delivery technologies,” she comments.

In fact, Konagurthu believes that AI/ML technologies “are poised to further accelerate the journey from molecule to medicine—leading us to a healthier society.” He adds that “as the industry continues to evolve, embracing these technological advancements will be key to overcoming long-standing challenges in drug development and unlocking new possibilities in the quest for more effective and accessible treatments.”

References

1. Jiang, J.; Ma, X.; Ouyang, D.; and Williams III, R.O. Emerging Artificial Intelligence (AI) Technologies Used in the Development of Solid Dosage Forms. Pharmaceutics 2022, 14(11), 2257. DOI: 10.3390/pharmaceutics14112257
2. Allen, C. Machine Learning Directed Drug Formulation Development. Adv. Drug Del. Rev. 2021, 175, 113806. DOI: 10.1016/j.addr.2021.05.016
3. Dong, J.; Wu, Z.; Xu, H.; and Ouyangt, D. FormulationAI: A Novel Web-based Platform for Drug Formulation Design Driven by Artificial Intelligence. Briefings in Bioinformatics 2024, 25(1), bbad419. DOI: 10.1093/bib/bbad419
4. FDA. Artificial Intelligence and Machine Learning (AI/ML) for Drug Development. March 18, 2024. https://www.fda.gov/science-research/science-and-research-special-topics/artificial-intelligence-and-machine-learning-aiml-drug-development.
5. Richards,K. U of T Scientists Use AI to Fast-track Drug Formulation Development. Press Release. Jan. 11, 2023. https://www.utoronto.ca/news/u-t-scientists-use-ai-fast-track-drug-formulation-development.

About the author

Cynthia A. Challener, PhD, is a contributing editor to Pharmaceutical Technology®.

Article details

Pharmaceutical Technology®
Vol. 48, No. 6
June 2024
Pages: 14–17

Citation

When referring to this article, please cite it has Challener, C.A. Using Advanced Algorithms to Solve Formulation Challenges. Pharmaceutical Technology 2024 48(6).