Meeting Highlights: DIA Global Annual Meeting 2023
Revolutionizing the Life Sciences:
How Diversity, Innovation, and Artificial Intelligence are Accelerating the Future of Health
“

It is not the strongest of the species that survives, nor the most intelligent, but the one most responsive to change.”
—Charles Darwin

Technology has helped bring us to the cusp of several crucial conversations in which stakeholders across the entire life sciences and health ecosystems must engage. Diversity, innovation, and AI can create a hugely powerful synergy that drives transformation across the health and life sciences. But clever ideas and inventions are no longer enough;—for these innovations to breathe new life into the real world, we must also work to create an environment ready to receive and adopt them.

AI: Accelerator on our Road to Transformation

Innovation is the science of transformation. Innovation empowers us to create new products, services, and processes that can improve the quality and efficiency of the health and life sciences. Innovation can take many forms: not only medicines or devices but also the communications and services accompanying their use. Regardless of our individual roles—as regulators, innovators, investigators, physicians, patient advocates, and others—innovation in its broadest sense must ultimately include the value of that innovation to the public.

Diversity and its counterpart inclusion are the fuel of innovation. They enable us to generate novel ideas that can address the complex and evolving challenges of health and life sciences in the gloriously diverse real world. But inclusion is also a moral obligation upon us all. It’s a critical imperative for us to think about diversity from multiple perspectives, not only the diversity of the people we’re working to serve but the diversity of the teams with whom we work to serve them. Leveraging the real-world lived experience of all the different people around us is the only way to investigate and articulate society’s most challenging problems, so that our solutions deliver the most important innovations that benefit all humankind.

ChatGPT, Microsoft’s Bing, and other recent generative AI innovations are demonstrating that AI can help us retrieve and augment insights, optimize processes, and generate customized outcomes or results. Artificial intelligence will not be a replacement for human intelligence, but instead a partner that can enhance our capabilities and creativity. AI is an accelerator on our road to transformation.

“There is one aspect that you could start using today. It’s too early for some things. But our administrative processes: Remember, in medicine, a third of our costs in the United States are administrative, billing, reimbursement. That’s easy pickings for AI. There are several administrative processes in clinical trials and target selection. I took a consent form from the web and told GPT to be the “front end” of that consent form. A patient can ask the consent form: If I’m in the control arm, will I get an injection? If I’m in the control arm, do I get any benefit? What’s the likelihood that I will get medication? All those questions that we never have time to answer now can be answered at the pace of the inquirer.”—Isaac Kohane
We innovate by the ideas we generate. To be innovative in science, and the associated use of data science, AI, machine learning, and real-world evidence, from target identification all the way through regulatory approval and reimbursed patient use, a diversity of thoughts and perspectives is required. The best ideas are generated by teams with diverse backgrounds, experiences, and thoughts, and so we must be intentional in building teams that are diverse and inclusive. This intention requires a leadership commitment to a mindset where true diversity is not just an aspiration but the standard way of doing business.
If you really want to transform, evolve, illuminate, etc., in an organization and in how you make medicines, you have to do it from the center. You have to incorporate new ways of doing things into your strategy, in how you’re making decisions, including how you’re designing programs and how you’re executing them. You also need diversity across your team and your talent. Statistics show you get better results when you have diversity on your leadership team and throughout every single level of your team, and one big reason for this is because innovation is driven by the ideas we generate, and you want ideas to be generated by those with diverse backgrounds, diverse experiences, and diverse thoughts. In our R&D data science organization, nearly half the team are women in an industry where the average is in single digits—and we have people from over 30 countries. That doesn’t happen by accident. Being diverse and global in our mindset plays a really important role in how you pick the questions you want your research to answer.—Najat Khan

Be Bold Enough to Play

To capture some of this opportunity, we must be bold. We must invest time and thought into playing with these tools to help imagine what the future could hold if we could unlock their true potential.

AI-based tools have the capability to be our co-pilot in everything from writing email to scheduling; they can play a role in uncovering disease-driving mutations, identifying novel disease targets, designing molecules, integrating real-world data, enabling decentralized trials and research, identifying clinical trial sites, principal investigators, and study participants, and more.

I’ve noticed that the people who’ve been closest to this technology have been the most excited, and yet also terrified. The people closest to it are both scared and excited. Other people have no idea: They think this is just a Google and don’t understand it yet. This is unlike any other technology in terms of its uptake. Our industry has a history of being extremely slow because we are more regulated. Everyone can think of many things that took way longer than we thought. This is just going to be everywhere. If you want to be in this industry, this is not the technology you can ignore. You can’t sit on the sidelines. You need to learn how to do a good prompt. Get engaged. Although the whole thing is hyped up, in some ways it’s not hyped up enough.—Amir Kalali

Moving Toward Efficiency and Transparency—and Beyond

Increasing efficiency in the life sciences has enormous value, and many processes can potentially be sped up by utilizing AI-based approaches.

Fueling innovation also means equitable access to research. We are already seeing how remote tools and decentralized methodologies can help ensure more inclusive trials. Many groups, including DIA and the Decentralized Trials and Research Alliance (DTRA), are collaborating with patients, clinicians, data scientists, operations leaders, regulators, and others to develop the frameworks, training, roadmaps, and effective communication to both patients and clinicians that foster best practices for technology-powered remote and decentralized research across this ecosystem.

Challenges of Translation and Extrapolation

On the other hand, extrapolating from the amazing linguistics and primacy of the large language models (LLMs) to what they can do with our data is not a direct translation. Life sciences data is very focused around individual questions. It does not support answering all questions.

We can go farther by working together and thinking through the complexities of what data is needed and what is the most efficient, inclusive, equitable way to collect that data. For example, what data should be developed pre-competitively, and then competitively, to get the full benefit of these foundational models? How we approach questions like this will determine whether we will be able to move beyond our current state to answer bigger questions in the life sciences using AI.

Data itself can present another challenge. For example, much of the data informing large language models was not originally generated or intended for research, so we would be wise to proceed with caution when using these models to generate content for this purpose. We would also be wise to recognize the possibility of inherent bias if the data does not represent the full population spectrum, which is often the case. We must remain aware of which data we are using and which questions it can (and cannot) answer.

To truly innovate in the application of technology, start with both problems and solutions. Understand where the challenges lie and where the innovation stands and investigate permutations of both to identify opportunities to truly innovate.

For example, the application of AI tools such as ChatGPT to summarize information has already begun. This can result in the tool generating hallucinations, things that are partially or completely inaccurate or untrue. We might consider such hallucinations to be mistakes—and we certainly need to be cautious about this. But they can also prompt us to consider ideas that we might not have otherwise thought about. Properly employed, AI tools can help make connections from where we are now to where we might potentially be in ways that are impossible for humans to imagine without them.

I happen to be the principal investigator of the Undiagnosed Diseases Network Coordinating Center. It’s led out of Harvard but involves 12 academic health centers, coast to coast. Until very recently, we had a very successful process where we would evaluate patients with a team and then sequence their genome. Then, based on a small number of candidate mutations, we decide which one of these are the likely candidate, and then do a CRISPR knock-in into a model organism and say, that’s the cause of this individual’s disease. We’ve helped literally hundreds of individuals, having evaluated about 35% of the individuals referred to us.

Back when I was first introduced to GPT4 in October 2022, I ran one of the cases that we had failed to diagnose by it. We had five gene candidates but none of the clinical findings were typical. I just fed the whole case into GPT4 and asked which is the one causing disease? It picked one. It sounded reasonable. But meanwhile, we were also doing the bench work to validate. And it so happened, for exactly the reasons it articulated, that was the mutation.—Isaac Kohane

In short, technology—closely coupled with scientific knowledge and operational know-how—can help us discover insights into deeply rooted challenges and find corresponding opportunities for truly meaningful change.

But it is ultimately up to us to steer and drive these changes and use these tools for a higher purpose.

The ability to fully leverage the opportunities before us still comes down to us: our people, our talent, and our willingness to engage, adapt, and change mindsets.

“If we go back 140 or 150 years ago, there’s an analogous situation. The internal combustion engine was just invented. But people didn’t know if cars were going to be popular or not. And some people were saying cars are more dangerous than dynamite; or don’t buy cars because they require maintenance; or horses are more predictable, and so on. We’re in a similar situation. We should strive to be the drivers of, not the passengers in, these new cars. Some people are going to be car mechanics. We all don’t have to be mechanics, but I hope all of us can be drivers. The clutch might be finicky, the transmission might still be manual. But if we don’t do it, then someone else is going to do it for us.”—Armen Mkrtchyan
This article summarizes the DIA Global Annual Meeting opening plenary discussion featuring Amir Kalali (Decentralized Trials and Research Alliance, DTRA), Najat Khan (Janssen), Isaac Kohane (Harvard Medical School), and Armen Mkrtchyan (Flagship Pioneering), moderated by Junaid Bajwa (Microsoft, UK).
Back to Issue