Search
Close this search box.

How AI is Changing Early Drug Development

Ramji Vasudevan, Senior Engineering Leader, Altimetrik

Innovations in drug development lead to life-saving medicines that improve and extend people’s lives. Bringing a new drug to market requires vast human insight, data, and rigorous testing to ensure safety and efficacy.

The payoff is substantial — the Food and Drug Administration (FDA) approved 55 drugs in 2023,1 the second-highest total in the last 30 years — but the road to getting that approval is long.

According to the Pharmaceutical Research and Manufacturers of America,2 it takes an average of 10-15 years and $2.6 billion for a new drug to journey from initial discovery to approval and widespread patient access. The massive investments of time and money create strong incentives for pharmaceutical companies to leverage technology to accelerate the drug development process.

Generative artificial intelligence (GenAI) offers the potential to reduce lead times, particularly for lengthy processes like defining clinical trial processes and protocols. Pharma companies are still in the early days of implementing such solutions, but progress is being made in reducing time to market.

One of the most time-consuming elements of that journey is clinical trials and early research and development (R&D). Pharmaceutical companies must design protocols and procedures, recruit institutions to conduct trials at contracted sites, enlist patient volunteers, collect, and analyze resulting data, monitor safety, and prepare trial reports.

All these steps, and the friction between them, typically cost the industry tens of billions of dollars every year and delay time to market. With a more coordinated and thoughtful approach to compiling all the necessary information for drug development, pharma companies can accelerate some of these processes and reduce the time and monetary costs associated with delays.

Even a task that seems easy, like getting data to the FDA during a clinical trial, typically takes four to six months after the patient’s last visit. With AI, we can reduce that wait to mere days.

Getting the data house in order

While AI promises accelerated timelines, successfully leveraging algorithms first requires building a solid data foundation. Integrating disparate data distributed across clinical trial sites, patient health records, genomic databases, and internal sources is the best path forward.

Centralizing data into a single source of truth enables standardization, governance, analytics, and the application of machine learning models. Many leading drug companies have invested in infrastructure, creating data lakes and data mesh architectures to bring together structured, unstructured, and streaming data sources.

Establishing trusted data with pedigree is even more vital than infrastructure modernization on the journey to AI augmentation. Documenting how data is collected, where it resides, if it needs cleaning, and how accessible and accurate it remains over time ensures quality. With inconsistent data, any downstream AI risks generating misleading outcomes.

Building a robust data posture isn’t a one-person job. It’s about developing a culture with process elements that go beyond technology. Data literacy training to democratize insights, empowering roles like data stewards, and even appointing chief data officers can help organizations that seek to implement AI and make evidence-based decisions. With the right people, processes, and performance metrics reinforced, AI adoption follows more seamlessly.

Opportunities for AI acceleration

Once organizations have done the critical work of preparing and consolidating their data assets into high-quality, trusted sets, they can apply AI and machine learning to advance their efforts. This is where the real payoff enters the equation.

However, no single flip-the-switch AI solution can optimize the entire drug development value chain overnight. Pharma companies need to identify where injecting intelligence can provide the greatest return and work steadily to connect those point solutions.

Some of the areas showing promise for initial AI improvement are:

  • Clinical trial protocol design: GenAI models can rapidly synthesize patient data, research archives, and past trial data to produce first drafts of trial protocols automatically. This can shave weeks or months off this lengthy documentation process.
  • Patient cohort selection: Algorithms can predict which patient demographic groups have higher adherence rates over the duration of trials. Ensuring cohort retention pays dividends.
  • Manufacturing planning: AI forecasting helps balance drug production batches to avoid shortages or waste.
  • Regulatory submission writing: Language models hold intriguing potential to auto-generate regulatory documents using historical submissions as examples, though no company has fully automated reporting end-to-end yet.

The key is to start with a few bounded, high-impact areas for AI rather than overhauling complex workflows all at once. Over time, assembling these modular improvements begins compounding efficiencies across the drug development value chain.

How to take the plunge

A major pharmaceutical company sought to become a top-five global leader in drug development and clinical trials. As a first step, it needed to benchmark processes and infrastructure against best-in-class standards. Bringing a new drug to market involves navigating a complex, interconnected web of processes across many stakeholder groups.

To accelerate its R&D velocity, this pharma company conducted an intensive six-week assessment of its end-to-end workflow. Over 50 interviews with subject matter experts mapped out 36 distinct sub-processes and scored maturity levels in areas like process friction, measurement, data utilization, and digitalization.

The analysis uncovered numerous process silos stemming from disconnected systems and information gaps. Compared to top industry peers, opportunities existed to optimize through better KPIs, data integration, and platform modernization.

With tangible insights on friction points and roadblocks hindering speed to market, the company can make strategic changes. The benchmarking study provided an honest look at internal challenges while offering a prioritized roadmap to drive efficiencies.

While no pharmaceutical company has yet achieved the full promise of AI to accelerate time to market radically, meaningful strides are being made across the drug discovery funnel. Whether optimizing preclinical molecular modeling, automating clinical trial protocols, improving manufacturing forecasts, or accelerating regulatory submissions, AI drives incremental efficiencies. Extending these modular improvements into an integrated platform remains the industry’s vision.

With sound data practices and governance in place, AI and generative models can accelerate each stage of the drug development journey — from early R&D through clinical trials and eventual commercialization. The future of getting life-saving medicines to patients faster through technology remains bright, even if adoption moves progressively rather than overnight.

Ramji Vasudevan is the Senior Engineering Leader at Altimetrik

References

  1. https://www.nature.com/articles/d41573-024-00001-x?_hsmi=289063790&_hsenc=p2ANqtz–2cbOmw5HlGnAD8BR8TXzlqVIrV1fl-OUQAulUuBPlizQPXcM45yGrTd4Q0pfqaLkqMQNEYyJ_BNMY0ZNVEqR44VAhWQ
  2. https://phrma.org/policy-issues/Research-and-Development-Policy-Framework