The journey to developing a successful drug, theoretically, may appear linear: you discover the right drug, find the suitable patient and administer it at the right time. The reality, however, often deviates from this straightforward path. Aligning these three variables remains notoriously difficult, often leading to elongated timelines strewn with failures, sometimes extending over a decade with costs often in the billion-dollar range.

In recent years, the use of AI in drug discovery and development has grown swiftly, marking a significant shift in how we understand, discover and develop new drugs. The technology promises to chip away at timelines and save the industry billions of dollars eventually. But those promises aren’t exactly new. Even before ChatGPT became popular, many drug developers were working with high-end algorithms to “try to really hone in on databases,” said Andrew Strong, a partner at Hogan Lovells.

Overall, the industry remains at a relatively early phase of AI adoption. Yet recent AI advances have elevated the process of querying databases for drug discovery applications. This advance enables a dynamic and iterative exploration of data where the system’s relational aspects can build on each other based on the questions posed. One caveat remains: Access to reliable, curated databases is a prerequisite. Without it, conducting searches to enable accurate and efficient data exploration for drug discovery is impossible.

AI’s growing role in drug discovery and development

Fellow Hogan Lovells partner Barry Burgdorf expands on this point, noting that AI can be instrumental in sifting through past clinical trial data to optimize current drug development processes. “AI can figure that out,” he said. “There are vast repositories of past clinical trial data that can be parsed to tweak approaches to see if something will work better.”

AI can not only help design molecules for specific indications but also help with patient selection for clinical trials. In that domain, AI can aid in identifying suitable candidates and inform clinical trial design. For instance, Boston Conulting Group notes that the AI-fueled pipeline is growing at an annual rate of nearly 40%, and AI/ML technologies are simplifying an array of drug development and repurposing processes.

In addition, AI can facilitate clinical trial enrollment, “which is a huge issue,” Burgdorf said. “There are drugs waiting to get into trials but they can’t enroll enough of the right patients.” He added, “AI can help find those right patients by analyzing data to predict who will respond best to a drug and should be enrolled.”

Strong further noted, “One area we’re seeing growth in is clinical trial networks — taking research institutions and putting them in networks with shared platforms to enhance trial capabilities and share data. As AI gets better at harvesting data, people will look for ways to share it while respecting patient privacy.”

But AI’s potential isn’t confined to early-stage drug discovery and clinical trials, Burgdorf and Strong noted. The technology has the capability to impact the entire pharmaceutical value chain. Evidence of this comes in the form of the first drug fully generated by AI, INS018_055 from Insilico Medicine, entered clinical trials in 2021. It recently graduated to phase 2.

The CEO of Insilico, Alex Zhavoronkov, called traditional drug development a “molecular casino,” in an interview with The Guardian. With a success rate post preclinical studies of around 10%, Zhavaronkov estimates that AI technologies could at least double the success rate to 20%, potentially saving billions of dollars spent on drug development. In an interview with Drug Discovery & Development, Zhavaronkov described the promise of AI to ‘imagine the perfect molecules’ for drug targets.

Applications of AI across the pharmaceutical value chain

The potential of AI in drug development extends beyond theory into practical application. AI techniques can not only replace some animal studies with in silico trials, but they also have profound implications for the role of biomarkers in the industry. “If a biotech startup is designing a clinical trial and wants funding, one thing investors want is biomarkers to show the drug’s mechanism of action and potential for success,” Strong said. “AI’s ability to analyze drugs could suggest associated biomarkers, making it a valuable tool in this context.”

AI’s role also extends to post-marketing diagnostics, where it can provide predictive modeling. “If you look at every stage of drug development, from a scientist having an idea, collecting IP, developing a formulation, doing preclinical and clinical work, and marketing and distributing the drug — AI can play a role at every step if used properly,”  Burgdorf said.

Strong said he’s seeing more institutions, including academic medical centers like, “recognize the value of data rights tied to their clinical trials, as a lot of these trials are funded by federal bodies like the [National Institutes of Health] and the [National Science Foundation].” He continued: “We are entering an interesting period where institutions may start looking for returns on their investment in data.”

Contract research organizations (CROs) are also beginning to explore how AI can be integrated into their services, according to a Boston Consulting Group report. AI-focused drug discovery companies are forming partnerships with CROs and other industry collaborators to co-develop and commercialize assets. This ecosystem approach has the potential to accelerate discovery, allowing AI firms to focus on their differentiating capabilities.

Privacy implications of AI in drug discovery and development

While AI brings vast potential to accelerate drug development, expanded use also raises new challenges around data privacy. As Burgdorf explains, “One area we’re seeing growth in is clinical trial networks — taking research institutions and putting them in networks with shared platforms to enhance trial capabilities and share data.” As AI gets better at harvesting data, professionals will look for ways to share data while respecting patient privacy. “Over the past 20 years we’ve become concerned about personal medical data privacy — HIPAA, data standards, security protocols,” Burgdorf said. “So there are barriers to open data sharing, even though it could help discovery.”

Strong also cited an example of data collection, “Argenx has a drug, cusatuzumab, licensed out to Johnson & Johnson,” he said. “During an AML trial, they gathered comprehensive omics data from the patients, forming a three-dimensional database that can be explored in multiple ways.”

Yet while the potential benefits of data sharing for AI-driven drug discovery are significant, so too are the patient privacy worries. These aren’t the only concerns requiring careful navigation as the industry integrates AI further into its operations.

Strong emphasizes the need for caution in integrating AI into healthcare, “AI should not be viewed as practicing medicine, but rather as a tool delivering results, not making treatment recommendations. Ethical issues may arise if AI is perceived as providing medical advice, which we need to avoid.”

“AI is unlikely to replace medical scientists soon,” Burgdorf agreed. AI is also unlikely to replace human doctors, as some industry observers have predicted. “We’re far from accepting treatments solely based on AI’s suggestion without the application of human medical judgment.”

---

Filed Under: clinical trials, Data science, Drug Discovery, Drug Discovery and Development, Industry 4.0, machine learning and AI
Tagged With: AI and pharma, biomarkers, clinical trial networks, clinical trials, data-driven drug development, drug discovery, patient privacy
 

---