In a rapidly evolving landscape where artificial intelligence (AI) and machine learning (ML) have become household terms, their impact on biopharmaceutical research and epidemiology is profound. I caught up with Mike Munsell, PhD, Director of Research at Panalgo, to discuss how these technologies are transforming the healthcare research landscape and their real-world applications in biopharmaceutical R&D. Classically trained epidemiologists know more about ML and AI than they think they do, and here’s why.
Q: We’re hearing more about AI, ML, and ChatGPT these days. Can you give me a definition of what each is and how they might overlap?
A: With AI, computers are essentially mimicking human decision-making. ML, a subdiscipline of AI, uses algorithms or statistical models to read patterns in the data to make a prediction. ML generates outcomes based on training data and makes predictions using this historical information. It’s up to researchers to interpret these predictions, and determine whether to leverage the information they suggest or set them aside.
ChatGPT relies on ML algorithms, but the model is so complex that it can generate clear, concise recommendations based on text data and prompts. ChatGPT learns not just the text, but how people talk. This is a good example of AI; when a network of trained ML algorithms can interact with a user in a way that we generally only associate with humans.
Q: When did big data, AI, and ML really explode onto the landscape, and why?
A: AI and ML are more established in the consumer sector, as retailers like Amazon and CVS have been tracking consumers’ buying habits and making recommendations for decades. Now, AI and ML are impacting the biopharma sector in a bigger way.
The advent of electronic health records resulted in much larger datasets, and over the last 10 years, different ways to capture data have become more widely available. Big data is the automation and digitization of everything, and the healthcare sector is a growing part of that evolution. To generate evidence and draw meaningful conclusions, you need advanced methods like AI and ML, which are designed to handle the unique characteristics of big data in ways that traditional statistical methods are not.
Q: What are some recent examples of positive use cases for AI/ML in biopharmaceutical R&D?
A: ML is being used in real time to aid doctors at the point of diagnosis or care. In drug discovery, ML helps clinicians understand how certain molecules might behave in a laboratory and assigns them probabilities on whether the molecules will be effective and safe. All this information can help researchers develop drugs faster and make more informed decisions. ChatGPT can make reading case reports for adverse events (or combing the literature) more efficient. Automating processes for researchers and clinicians will be AI’s primary use factor in biopharma in the near term.
Q: On the other hand, are there any cautionary tales around AI/ML in biopharmaceutical R&D?
A: People are worried that AI will replace human decision-making, but the biopharma industry should instead consider AI an “efficiency copilot” that automates time-consuming tasks: it helps with diagnosis and treatment decisions and predictions, but it doesn’t replace human decision-making. A clinician or researcher is still in control of the actions taken based on these recommendations.
Q: Do epidemiologists and data scientists approach their research in different ways? Let’s talk about both common and unique ground.
A: Epidemiologists are more focused on the “big picture;” they tend to have a more holistic view of the research question at hand and what they want to learn from the data, whereas data scientists may be more focused on specific tasks, algorithms, and data behavior that can aid in answering that research question. Epidemiologists and data scientists should work together to maximize AI’s full potential. Epidemiologists need to understand AI/ML methods, and they can rely on trained data scientists or a software solution to execute these methods. Most epidemiologists have already used many of the AI and ML models that could be used to inform their research. A logistic regression, for example, is the simplest form of ML.
Q: How does Panalgo support all researchers across the spectrum of RWE generation?
A: You can analyze data in our robust and rapid IHD platform, or we can provide a trained analyst to help with your project work, creating the bundle of data, services and technology that’s right for you. Our platform has applications for the whole spectrum of data exploration, hypothesis generation and testing, and predictive modeling.
Ready to enhance your analyses with the power of machine learning? Contact us today to find out how Panalgo’s IHD Analytics can work for you.