Vasan Yegnasubramanian, MD, PhD, is the director of Precision inHealth Medicine at Johns Hopkins. He spoke with Johns Hopkins Kimmel Cancer Center communications staff about how AI is transforming cancer research and treatment. The transcript of the conversation follows.
Why is AI important in cancer research and treatment?
Artificial intelligence (AI) is poised to fundamentally transform health care. It’s not just about adding a new tool—it’s about rethinking how we deliver medicine. Data is a valuable asset, but it’s only powerful when refined and used wisely. AI helps us do that. It enables predictive, real-time, and consistent care and opens the door to innovations we couldn’t even imagine before. Think about early cancer detection, individualized treatments, and accelerated drug discovery. AI gives us the potential to change everything, but we must do it responsibly and ethically.
You mentioned four main priorities. Can you walk us through them?
I often compare data to oil. Just like oil, it’s incredibly valuable, but not in its raw form. Priority 1 is ingestion: figuring out where the high-value data assets live and pulling them from their source systems, like medical records or imaging databases. Data sitting in a medical record helps us deliver clinical care, but it doesn’t help us innovate. We have to extract it.
Priority 2 is annotation and curation: structuring, cleaning, and organizing the data so it can be used. Most of it isn’t ready for research in its current form. This process is like refining oil, turning raw material into fuel for innovation.
Priority 3 is building distribution channels and computing platforms. Rather than copying and sending massive data sets to researchers, which is risky and inefficient, we bring researchers and their tools to the data, where it securely resides. This is critical for both privacy and performance.
Priority 4 is responsible stewardship: We can’t allow unbridled access to sensitive data. There is a lot of power here, and we must build in governance to ensure that our innovation doesn’t cause harm. At Johns Hopkins, we established an AI and Data Trust Council to help oversee this.
How else are you addressing concerns around privacy and public trust?
We take security very seriously. Security is actually most vulnerable when data is decentralized—sitting on individual computers. A bad actor or hacker only needs to breach one machine. That’s why we’re shifting to centralized, monitored systems where access is logged, and unusual activity is flagged immediately. We also have a Chief Information Security Officer, who helps us identify and prevent potential breeches and is putting systems in place that generate an immediate response if a breech occurs. Our goal is to share as much data as possible for the benefit of research while ensuring that patients remain anonymous and protected.
What does it mean to be a steward of public trust in this context?
Much of the data we use was collected for clinical care. Now, we’re repurposing it for secondary research. We have legal frameworks for that, but more importantly, we have an ethical obligation to show that what we are doing is truly for society’s benefit. The public must be a partner in this. That means being transparent, thoughtful, and inclusive. We make progress by building trust.
What can science enable with this approach?
This is where it gets really exciting. With interdisciplinary collaboration spanning medicine, data science, and engineering, we’re not just making tweaks, we are transforming health care. Here’s how:
- We’re shifting from reactive to predictive care. AI tools can help detect disease earlier, allowing us to act before symptoms become an illness.
- We’re turning ad hoc care into real-time care. AI agents can monitor wearable and smart device data 24/7, flag issues, and alert care teams in real time.
- We’re reducing unwanted variation. Whether it’s differences in access, knowledge gaps, or errors, AI helps ensure that best practices are consistently applied.
We’re accelerating innovation. AI can detect patterns in imaging that the human eye cannot. It can screen billions of virtual drug compounds, design new ones from scratch, and drastically reduce development timelines, and this is only the beginning.
How do these innovations help individual patients?
What we’re building is designed for both scale and personalization. The data sets we use are broad and de-identified, but the discoveries we make feed directly back into individualized care. That’s the essence of a learning health system—data from many helps the one. It’s a full-circle model where innovation informs better care.
What about concerns that AI might lead to overdiagnosis or overtreatment?
That’s a real concern, especially with early detection. Just because we can find something early doesn’t always mean we should treat it. AI can help us pair early detection with risk stratification—figuring out whether what we found is likely to become dangerous. For example, in prostate cancer, we’re developing biomarkers to distinguish aggressive from indolent cases, so patients get the right level of care. We will have similar strategies for every cancer type.
How is this getting back into the health care delivery system?
We’re developing tools like Patient Insight—a dashboard that helps doctors see a patient’s journey in context of thousands of others. Another project, the Health General Reasoner, is a backend tool to embed AI into clinical workflows. We are piloting a collaboration with Microsoft of a tool developed inhouse at Johns Hopkins that helps identify patients at risk for blood clots, a common risk of hospitalization, and suggests preventative treatment in real time.
Do you have plans to expand what we’re doing with AI beyond the Kimmel Cancer Center?
No single institution has all the data or expertise. That’s why we’re part of a national initiative called the Cancer AI Alliance with four other top cancer centers. We’re working to build a federated learning platform where each center keeps its data secure, but shares its models. AI tools travel to the data, not the other way around. This preserves security and privacy while enabling broad collaboration. If we succeed, it could serve as a model for every cancer hospital in the country.
Is Johns Hopkins is helping to lead the way?
In many ways we are, but we’re not doing this alone. We’re working across institutions, with industry, and with our communities. The technology is powerful, but it’s the people and partnerships that will make it truly impactful.
The future is incredibly promising. We’re exploring how AI can accelerate drug development, personalize treatments, and help us understand cancer in ways we couldn’t before. But we’re also focused on making sure the technology is inclusive, equitable, and trusted. If we do it right, AI will raise the standard for everyone. That’s our mission.
More on the web:
Podcast with Vasan Yegnasubramanian:
https://cancer-matters.blogs.hopkinsmedicine.org/2024/12/19/cancer-matters-bill-nelson-artificial-intelligence/
Pancreatic cancer:
https://www.hopkinsmedicine.org/inhealth/pancreatic-cancer
Prostate cancer:
https://www.hopkinsmedicine.org/inhealth/prostate-cancer
