Pharmacogenomics (PGx), the study of how genetic profiles affect a person’s responses to medications, has already begun to help healthcare providers (HCPs) optimize their care by increasing the effectiveness of medications to prevent, undesirable ones Minimize side effects and improve patient experience. This rapidly growing field brings together bioinformatics and pharmacology and represents a transformative new era of precision medicine and highly personalized treatments, serving patients by helping physicians better predict therapeutic responses and more accurately optimize drug dosages.
However, with data-driven solutions come data-driven challenges, not least the size and complexity of the datasets on which pharmacogenomics relies. The enormous amount of genomic data and patient responses to medical treatments requires enormous human effort to analyze, and because distinguishing meaningful patterns (signal) from irrelevant data (noise) is such a challenge in large-scale data analysis, researchers may miss important connections between genetic information and drug reactions of the patient.
AI accelerates PGx insights and expands possibilities
AI has the potential to help PGx overcome its data analysis challenges because of its ability to efficiently analyze huge data sets and identify patterns and correlations that might otherwise remain hidden, helping researchers and manufacturers to help produce new, more effective drugs. Similar to how AI is used in industries like aerospace for predictive maintenance (e.g. analyzing engine data), AI systems in healthcare can be great at cutting through the noise; That means distinguishing normal genetic variations from those that indicate disease or predict drug reactions – a process for human researchers akin to looking for a needle in a haystack. But AI-controlled PGx systems can also help patients directly. By using the genetic profile data of their patients. Physicians can better predict individual responses to specific medications, helping to make informed treatment decisions that lead to better patient outcomes.
AI-driven systems can also use patient data to create digital twins – simulations of a patient’s physiological state – which can then be used to test different treatment strategies and gain new insights from individually tailored drug interaction data. This technology allows doctors to replace the traditional trial-and-error approach of many medical treatments with better, more personalized plans that can achieve better results. For chronic diseases like diabetes, the flexibility of digital twin technology also means providers can monitor, manage and predict how lifestyle and medication changes may affect things like blood sugar levels, making personalized treatment plans more adaptable and responsive to patients.
Challenges of AI-driven pharmacogenomics
However, despite its potential, AI in pharmacogenomics faces major challenges. Because datasets of genomic information and individual patient responses to medications are so large and so widely distributed across a variety of research platforms, electronic health record systems, and laboratory information management systems, integrating traditional PGx tools with the data to produce reliable insights is meaningfully difficult.
Healthcare professionals seeking to integrate pharmacogenomics systems into their practice also face significant resource challenges themselves. While affordability of tools and labor costs for implementation are always top of mind, providers’ internal need for the requisite genomic expertise to derive clinically relevant, actionable insights from these massive data sets presents a significant additional barrier.
Results-oriented AI tools
A variety of new AI tools have begun to address such potential challenges and demonstrate tangible results in PGx research and clinical applications while overcoming these barriers to data integration and vendor adoption. However, for HCPs choosing a tool, some differentiators are more important than others. AI-driven extraction tools, for example, used to interface with other electronic data systems (including electronic medical records) would be far preferred by physicians due to the resulting improvement in data integration and improved interoperability, especially if these tools were also more affordable than others the market.
The best new tools also leverage AI and advanced deep learning models to improve variant calling accuracy. Variant calling is the process of distinguishing true variants from errors. Because pharmacogenes tend to have more complex genetic variations and need to be analyzed differently than typical disease-related genetic variants, the process is complicated for traditional PGx tools. However, the right AI models, trained on large, annotated genomic datasets and using established variant detection algorithms, are reliably better at variant calling and provide much more precise predictions for clinical applications.
Finally, a tool’s maintenance schedule – how the data is updated to further train the underlying AI – is also a key differentiator, and some new genome extraction tools can leverage consumer DNA testing and whole genome sequencing (WGS) through partnerships with genetic testing companies and laboratories , making them attractive candidates for HCPs. These tools can extract PGx data from WGS data, allowing them to expand their genetic services to PGx without collecting additional samples or developing additional tests. The result is the generation of robust clinical insights that can be implemented by healthcare professionals at the point of care, without the need for further expert analysis.
New frontiers in pharmacogenomics
Pharmacogenomics as a field is already beginning to revolutionize healthcare, both in the research providers rely on and the personalized point-of-care decisions they make with their patients. With the help of AI, the predictive capabilities of pharmacogenomics are even greater, and with the right tools, healthcare professionals have the potential to emerge from this industry-wide paradigm shift to create a new standard of care that is as precise and powerful as it is patient-centered.
Photo: Khanisorn Chaokla, Getty Images
Peter Bannister, DPhil, serves as UGenome’s Chief Product Officer for UGenome AI, a precision medicine tools company that enables personalized treatment and dosing for every phase of therapeutic development.
Alan Kohler, PhD, serves as director of strategic communications at UGenome AI.