Time to Innovate for Faster, Smarter Discovery and Clinical Development Cycles
Contributed Commentary by Josh Gluck
August 6, 2020 | We’ve learned so much living through these unprecedented times. One lesson is the need to mobilize quickly, provide effective testing and contact tracing, and accelerate the development of therapies and vaccines to protect the global community. The issue is not simply one of speed; it’s more complex than that. We must also enable smarter development cycles with innovative approaches that provide the most expeditious, cost-effective, and, ultimately, safest path forward.
That can only be done through the power of data: the defining currency in drug discovery and clinical development. We need to be able to leverage more of it, faster, and more completely than ever before to shave time off the drug discovery and development time frame.
Artificial intelligence (AI) is playing a growing role in nearly every phase, from compound identification to clinical trial recruitment to efficacy or toxicity detection during early or later-stage clinical trials. We’ve only begun to scratch the surface. The next frontier lies in operationalizing AI to ensure rapid iteration and expansion across the enterprise.
What can we do to move faster? Let’s start by looking at where AI is having an impact today.
Mining existing research
AI is gaining traction in automating the process of mining scientific research and patient data – saving massive amounts of time and reducing the cost associated with manual research. In addition, researchers are leveraging AI to explore data from thousands of existing drugs and compounds to determine whether they can be repurposed for new therapeutic uses. We already see applications of this in potential treatments for COVID-19, accelerating time to clinical trials, and hopefully, time to market.
Researchers are leveraging AI to mine data on various molecules, compare them against desired parameters, and identify the most promising compounds for synthesis and testing. Additional research in this field is emerging, with machine learning (ML) and quantum physics combining to deliver even greater impact. This has a ripple effect in accelerating subsequent stages of drug discovery, with fewer compounds needed for testing and development. One candidate, from British startup Excscientia, took less than a year from conception to being ready for trials.
Improving pharmacovigilance and efficacy
Safety is paramount in drug development and post-market. Pharmacovigilance remains a highly manual and increasingly expensive and complex process as channels for receiving data on potential adverse events continue to grow. Social media platforms are a prime example as they can contain important signals as to potential issues. We’re seeing AI technologies leveraged extensively in these areas. A new study by the Drug Information Association and the Tufts Center for the Study of Drug Development found that 57% of pharma and biotech companies are using AI for pharmacovigilance and risk management.
In addition, convolutional neural networking holds promise for optimizing potency and testing for off-target toxicity, as it enables companies like Atomwise to simulate and analyze molecular reactions and predict how they might act. The result is smarter development and identification of hits that provide maximum benefit and minimum adverse effect.
Simulating randomized control trials
The diseases and superbugs we find ourselves fighting today are incredibly sophisticated and complex. Despite massive investment in research and development, only 5 in 5,000 drugs reach market. We must move from correlation to understanding the structure by which a disease operates. For example, GNS Healthcare used AI to process genetic data of patients with Parkinson’s and uncover previously unknown mutations that may contribute to faster deterioration. Moreover, the company was able to create 5,000 computer simulations of randomized control trials—comparing a patient’s potential response to different drug scenarios. Simulations are a powerful development that ensures not just safety, but also optimizes R&D costs.
Creating a clinical data hub
Even though AI is advancing the drug development cycle with computerized trials, clinical trials with tests in animals and humans remain essential for Food and Drug Administration (FDA) approval. Researchers must document their results at each step—and report fatal or life-threatening suspected adverse events within seven days.
To address growing data requirements and complex data volumes, AI is already proving its usefulness in reducing the administrative burden associated with managing vast data sets associated with clinical trials. AI platforms that analyze biological and clinical data streams from diverse sources and formats—ranging from blood results to imaging—can also accelerate time to insight for investigators.
How organizations can create an infrastructure for AI success—faster
Recent events have underscored the need to move faster to seize the potential of AI technologies and methodologies. And yet, because AI is still a new type of deployment for many organizations, it presents unique challenges. They must consider where data will live, where workloads will run, and how to optimize their infrastructure to power analytics at record-breaking speed. Often, data science teams face operational challenges because they lack the oversight and automation to produce models efficiently, using repeatable processes. This limits the ability to operationalize ML at scale.
To achieve the full benefits of AI and extend into deep learning, organizations must realize a modern data experience and the foundation for it. A modern data experience enables data storage as a service that allows companies to extract maximum value from their vast amounts of data while reducing complexity and expense.
To scale quickly, some organizations are turning to joint ventures to leverage supercomputers. While these partnerships are feasible for well-resourced companies and make sense for short-term projects, many companies would benefit from having their computing power closer to their researchers’ fingertips. Data storage as a service makes this possible—allowing organizations to scale up or down quickly and cost effectively based on need. Above all, researchers need storage that is flexible based on usage, whether for mining research, identifying and evaluating compounds, or supporting clinical trials.
The scale of data available for investigators may also vary with each project. This is where smart provisioning becomes especially valuable. AI-based optimization enables life sciences companies to forecast where workloads should run for optimal business impact. It also helps optimize IT spend by forecasting data growth requirements, enabling leaders to make the most of R&D budgets.
Moreover, modern infrastructure must be able to power analytics at record-breaking speeds. Time is of the essence in discovering hits and developing life-saving therapies. To power AI—and faster insights—organizations must deliver low latency, high bandwidth, and maximum density.
AI applications must be supported by straightforward infrastructure, not complex setups. Infrastructure complexity is often the main reason for stalled AI initiatives and uneven adoption of AI for discovery and clinical development. In contrast, a fully integrated stack can accelerate the AI pipeline and avoid the need for intricate configuration and tuning so investigators hit the ground running. Simplicity remains the name of the game; the modern data experience enables researchers to seamlessly transition to a cloud model for maximum agility and improved efficiency at every stage of development.
We need to be able to leverage more AI, faster. The right infrastructure meets this urgency and is operational immediately, without slowing down any projects through downtime. Further, modern infrastructure ensures life sciences corporations remain on the cutting edge with continuous software and hardware innovation to support the latest AI advancements and computing requirements.
Josh Gluck is Vice President of Global Healthcare Technology Strategy at Pure Storage where he is responsible for Pure's healthcare solutions technology strategy, market development, and thought leadership in healthcare. He is also an Adjunct Assistant Professor of Health Policy & Management of NYU’s Robert F. Wagner Graduate School of Public Service. He can be reached at email@example.com.