How generative AI is revolutionizing drug discovery and development
Have you ever wondered just how long it takes to bring a new medicine to market? For those of us working in the pharmaceutical industry, the answer is clear: it can take decades and cost billions. As a computer scientist leading the AI team at AstraZeneca, I’ve seen firsthand just how complicated the drug discovery process is.
Trust me, there’s a lot of work that goes into developing a drug. But the real challenge lies in making the process more efficient and accessible, which is where generative AI comes in.
In this article, I’ll walk you through the critical role that AI plays in accelerating drug discovery, particularly in areas like medical imaging, predicting patient outcomes, and creating synthetic data to address data scarcity. While AI has incredible potential, it’s not without its challenges. From building trust with pathologists to navigating regulatory requirements, there’s a lot to consider.
So, let’s dive into how AI is reshaping the way we approach drug discovery and development – and why it matters to the future of healthcare.
Generative AI’s role in enhancing medical imaging
One of the most powerful applications of generative AI in drug development is its ability to analyze medical images – a process that’s essential for diagnosing diseases like cancer, which can be difficult to detect early on.
In the world of pathology, we’re increasingly moving away from using traditional microscopes, and using digital images instead. With digitized tissue biopsies, we now have access to incredibly detailed scans that show every single cell. The sheer volume of this data – sometimes containing over 100,000 x 100,000 pixels – makes it almost impossible for human pathologists to analyze every single detail, but AI can handle this level of complexity.
At AstraZeneca, we’ve been using generative AI to help analyze these vast datasets. One of the most exciting developments is in medical imaging, where AI models can quickly identify cancerous cells, segment different areas of tissue, and even predict patient outcomes.
For example, AI can help predict whether a patient will respond to a specific drug – information that’s invaluable for companies like ours, as we work to develop treatments that will provide real, tangible benefits for patients.
In my work, we leverage powerful AI techniques such as variational autoencoders and generative adversarial networks (GANs) to build these models. These AI techniques can help us learn from medical images and generate synthetic data that can be used to train AI models more effectively.
Marisa Garanhel
Overcoming data scarcity with synthetic data
In healthcare, one of the biggest challenges we face is data scarcity. High-quality, labeled medical data is incredibly difficult to obtain, and privacy concerns mean that it’s often not feasible to collect large amounts of patient data. This scarcity of data can be a significant barrier to building reliable AI models.
One of the ways we’re addressing this challenge is by using synthetic data. With generative AI, we can create realistic medical data that mirrors the patterns seen in real-world datasets. This synthetic data can be used to train AI models, allowing us to overcome some of the limitations posed by data scarcity.
In addition to providing data for training, synthetic data can help address another key issue in medical imaging: heterogeneity. Data collected from different hospitals or medical centers may differ in terms of image quality, scanning techniques, or even how the tissue samples are prepared. These differences can make it difficult for AI models to generalize and perform well across datasets from different sources.
To solve this, we’ve explored using generative AI to standardize the data. For example, if we train a model using data from one hospital, it may not work as well when applied to data from a different hospital.
By generating synthetic data that mimics the characteristics of different hospitals, we can ensure that our models work reliably across a wide range of real-world scenarios.
Marisa Garanhel
The role visual language models play in drug development
A particularly exciting area of research for us has been the use of visual language models. These models combine both visual data (such as medical images) and textual data (such as pathology reports) to create more comprehensive insights.
Imagine this: a pathologist receives a biopsy scan and needs to write a report that summarizes the findings. Instead of manually writing out every detail, a visual language model could analyze the image and generate an automatic report, outlining exactly what’s happening in the scan – what areas are abnormal, where cancer is present, and whether any other pathologies are visible.
In a project we worked on, we analyzed pathology reports from 14 hospitals across the UK, all related to chronic kidney disease. These reports were highly heterogeneous, with different hospitals using different formats, and the text was written in varying styles.
Using AI, we harmonized these reports into a standardized format, extracting only the most relevant information. This multi-modal approach – combining text and images – showed just how powerful generative AI can be in streamlining the process of medical data analysis.
Christie Mealo
Building trust with pathologists and medical experts
One of the biggest challenges we face in applying AI to healthcare is building trust with the medical experts who use our models.
Pathologists, after all, have spent years developing their expertise and have a deep understanding of the nuances of disease detection. Convincing them that AI can be a useful tool – and not a replacement for their expertise – requires careful consideration and collaboration.
From the beginning, we’ve worked closely with pathologists to ensure that the AI models we develop address their real-world needs. We don’t simply create a model and then hand it off to the pathologist; instead, we sit down with them, understand their challenges, and develop the model together. This collaborative approach ensures that the AI we build is both useful and trustworthy.
One example of this collaboration is our work on prostate cancer detection. The challenge with prostate cancer is that some cancerous glands are so small that even experienced pathologists can miss them. But AI can spot these tiny glands with impressive accuracy.
When we showed our model’s results to pathologists, they were impressed by how effectively the AI was able to detect these hard-to-find glands. This success then helped build their trust in the technology.
Ana Simion
Ensuring accountability and regulatory compliance
While AI can significantly enhance the drug discovery process, it’s important to remember that pathologists remain accountable for the final medical decisions. AI is a tool that supports their work, but it’s not a replacement for their judgment.
At AstraZeneca, we are deeply committed to ensuring that all of our AI applications comply with regulatory standards. The healthcare industry is heavily regulated, and we take compliance seriously. Every model we develop must meet rigorous standards before it can be deployed in clinical settings.
We also recognize that developing AI models for healthcare involves an ethical responsibility. These models must be transparent, explainable, and free from bias. It’s not enough for the model to work; it must also work fairly and reliably across all patient populations.
Conclusion
So, while AI is an incredibly powerful tool, it’s essential to remember that pathologists, clinicians, and researchers will always remain at the heart of drug discovery. AI is here to support their work, not replace it.
As we continue to advance in our use of AI, it’s exciting to think about the possibilities for improving patient outcomes and accelerating the development of new medicines. But, at the end of the day, the human touch – expert judgment, empathy, and care – will always be paramount in healthcare.
At AstraZeneca, we’re excited to be at the forefront of these changes, and we look forward to continuing our work with AI to make a meaningful difference in the lives of patients around the world.
This article comes from Dr Nikolay Burlutskiy’s talk at our London 2024 Generative AI Summit. Check out his full presentation and the wealth of OnDemand resources waiting for you.
Nikolay was the former Director of ML and AI at AstraZeneca at the time of this talk. He’s currently the Senior Manager of GenAI Platforms at Mars.