How AI Will Impact the Drug Discovery Pipeline 

Creating new medicines is a laborious process, requiring years of worth and mountains of cash to make any significant advances. With so much money and literal lives at stake, speeding up the drug discovery process is always a top-of-mind topic for industry professionals.

Alister Campbell Credit: Dotmatics

Like every other industry that involves a lot of time-consuming tasks, drug discovery is experiencing a revolution with the introduction of AI tools.

While human-driven trials will always play a role in the medical world, AI can help researchers narrow their focus to the potential drugs with the highest chances of success.

To learn more about this exciting field, we spoke with Alister Campbell. Vice-President and Global Head of Science and Technology at R&D software firm Dotmatics, Campbell has years of experience watching technology trends come and go. As you’ll see, he firmly believes AI has enormous potential for increasing the efficiency of the drug discovery pipeline.

How has the integration of artificial intelligence in drug discovery impacted the traditional timeline of bringing new drugs to market?

AI has the potential to revolutionize the pharmaceutical industry by significantly enhancing the typical drug discovery and development process. Traditionally, bringing a new drug to market has taken approximately ten to fifteen years and costs around $2.5 billion. However, AI is changing this landscape by shortening development timelines and reducing costs. Life science companies can now leverage vast amounts of data that were previously difficult to access. This data, once organized, allows for advanced analyses such as generating reports, conducting ad hoc queries, and creating interactive visualizations, which help scientists identify patterns within their drug discovery data.

AI has the potential to positively impact all aspects of drug discovery, but in particular the opportunity to speed up the early stages of drug discovery by quickly analyzing scientific data to predict and pinpoint promising candidates faster than empirical development.

This not only accelerates research, but also by marrying early research data with the data from clinical studies, AI could potentially improve the accuracy of predictions regarding the safety and efficacy of potential drug compounds. This could reduce the likelihood of costly failures at later stages. AI also enhances the design and participant selection for clinical trials, tying together phenotypic and genotypic data, making these studies more efficient and effective. Additionally, AI's capability to analyze real-time data from ongoing trials allows for immediate adjustments, optimizing the development process and resource management.

By integrating AI, the pharmaceutical industry is not only speeding up the drug development process but also driving innovation by gaining deeper insights into the complex biology of diseases. Think of AI like a powerful flashlight that illuminates hidden patterns and insights within vast amounts of data, allowing us to see and understand aspects of diseases that were previously too complex or obscure. This capability leads to the faster delivery of more effective therapies to patients. The use of AI in drug discovery promises a future where new treatments are developed more quickly and at a lower cost, potentially transforming patient care and outcomes.

What emerging AI technologies show promise for shortening the drug discovery process and potentially accelerating the development of new treatments?

We are only just beginning to understand what AI is capable of in terms of research and development. As AI continues to advance, it promises to unlock discoveries that are currently unimaginable. Today, AI’s potential is maximized by scientists at the decision-making juncture.  Providing high-quality, well-annotated, and reliable data to AI engines will allow, in real-time, AI to autonomously identify patterns within drug discovery data to generate predictions and even propose novel hypotheses for further investigation.

For example, Dotmatics has already developed some of these state-of-the-art AI applications within our own tools to assist in use cases such as the use of flow cytometry, and we’re working directly with customers to increasingly add such AI analysis functionality to new areas to support their needs.

This and other use cases can be aggregated into Dotmatics new cloud-based platform called Luma which aggregates relevant data across labs including lab instruments into clean datasets for analysis which paves the way for AI & ML-based algorithms.

What role is multimodal drug discovery playing?

Pharmaceutical and Biotechnologies are no longer married to any one entity type as a therapy for a given disease or target.  Multimodal drug discovery at its core is the ability for scientists to pick the best therapy or combination of therapies to address a particular target.  It involves researching and testing from across different domains of science in the process of discovering new compounds or therapies.

The most progressive players in drug discovery are accelerating toward an AI-enabled, multimodal drug discovery future. Example therapy modalities include antibodies & other proteins (including antibody-drug conjugates), cell therapies, gene therapies, RNA therapies, vaccines, peptide drug conjugates, and even within traditional small molecule therapies designs have evolved away from inhibitory or excitatory approaches to seeing the rise in targeted protein degradation (PROTAC’s).

How has Dotmatics been instrumental in supporting researchers to leverage AI for accelerating discoveries in drug development?

Dotmatics has built the world’s most powerful multimodal scientific discovery platform – Dotmatics Luma – connecting scientific applications, laboratory instruments, and other data sources to enable deep collaboration, automation, and analysis and to power an AI-assisted future. This R&D data management platform simplifies and accelerates the collecting and processing of instrument data and helps non-technical users easily gain critical insights directly from data.

Addex Therapeutics, a biotech headquartered in Geneva, Switzerland, is developing novel, orally available small-molecule drugs to address the unmet needs of patients with neurological disorders, such as Parkinson’s, epilepsy, Alzheimer’s, PTSD, depression, neurodegenerative conditions, and other central nervous system (CNS) diseases.

To quickly deliver such novel treatment options to patients with CNS diseases, the R&D teams at Addex must be nimble, data-driven, and collaborative, but doing such can be challenging when facing complex workflows, a true big data challenge with huge volumes, velocity and variety of R&D data, and occasional staff departures. Dotmatics has saved their scientists hours of time spent in analyzing data compared to our previous solution. 

What are some key challenges that researchers face when transitioning from traditional drug discovery methods to AI-driven approaches, and how are these challenges being addressed?

Transitioning from traditional drug discovery methods to AI-driven approaches presents several key challenges, including ensuring data quality and availability, integrating AI with existing workflows, maintaining interpretability and trust, navigating regulatory and ethical considerations, and managing cost and infrastructure needs.

A fundamental issue is that AI models are only as good as the data they learn from; bad data leads to bad science, emphasizing the need for high-quality, well-annotated data. Organizations are thus improving data collection and curation, and ensuring access to diverse datasets.

There is also a growing emphasis on documenting where AI/ML models are used and the training sets employed, which is crucial for accurately reflecting on model success in the future. Integrating AI technologies can disrupt established workflows, so institutions are gradually implementing AI tools and providing training to ease this transition. Interpretability remains a critical challenge; if a model is uninterpretable by an impartial and trained scientist, it can reduce trust in the model. Developers are enhancing the interpretability of AI models and conducting rigorous validation studies to build confidence.

As pioneers in scientific R&D software, we are committed to ensuring AI's responsible implementation, fostering honest discussions among stakeholders, and supporting scientists in harnessing AI's power to change scientific discovery for the better.

Additionally, the high costs associated with AI infrastructure and expertise are being mitigated through cloud-based solutions and collaborative models, like Luma Lab Connect allowing broader access to these advanced technologies. These strategic approaches are helping to pave the way for a smoother adoption of AI in drug discovery, aiming to make the process more innovative, efficient, and effective.

Any general thoughts/quotes on the future prospects of AI in drug research? What excites you?

As soon as this year, healthcare could begin to see the impact of the first AI-based drugs in clinical trials, illustrating just how AI-powered tools are revolutionizing patient profiling and diagnostics. This is just the beginning of reaping the benefits of digitalization, especially for life science companies that are increasingly adopting drug repurposing strategies.

As we've seen, AI's ability to navigate through vast, previously untapped data is not only accelerating the discovery of potential drugs but also re-evaluating compounds that are safe, but may have missed their initial clinical targets. This approach is particularly exciting as it offers a faster route to identifying effective treatments, leveraging the safety profiles of existing drugs to meet urgent medical needs.

Each new drug candidate, whether it succeeds in clinical trials or not, represents a step toward better health and quality of life for people with both common or rare diseases. Keep in mind that the introduction of ChatGPT wasn’t a Cambrian moment either. The idea of large language models dates back to the 1960s. Computer scientists and chip designers worked quietly and diligently for decades to make the release of ChatGPT possible.

Along the way, they delivered advances in data storage and processing that have transformed how we live and work. Pharma’s march toward the successful application of AI will be punctuated with the same small wins that add up to transformative change. It's a groundbreaking time in healthcare, where AI's integration promises to transform patient outcomes and drive a new era of innovation and efficiency in drug development.