mastering upset plots: a comprehensive tutorial for effective data visualization in pharmacometrics and drug development
mahmoud a. afifi (1), samer mouksassi (1)
Institution: (1) Integrated Drug Development, Certara, Princeton, NJ, United States.
Introduction/Objectives: In the realm of data analysis, the examination of variable intersections holds paramount importance as it often unveils crucial insights that are pivotal for informed decision-making across diverse fields, including but not limited to pharmacometrics specifically and generally to clinical research and healthcare management. While conventional visualization techniques, such as Venn diagrams, have been traditionally employed to represent these intersections, they frequently encounter limitations when faced with complex datasets, particularly when the number of sets surpasses a certain threshold, leading to what is commonly referred to as the "combinatorial explosion" problem. To address this challenge, recent advancements in visualization techniques have introduced upset plots as a promising alternative for effectively visualizing and interpreting complex data intersections (1). Unlike Venn diagrams, upset plots offer a more comprehensive and insightful representation of set intersections, particularly in scenarios involving a large number of sets or complex data structures (1). This tutorial provides a step-by-step guide to mastering upset plots, showcasing their practical applications through a series of case studies.
Methods: Each case study illustrates the utility of upset plots in different domains:
- Oncology: Understand how upset plots aid in exploring the pattern of pathology features associated with certain clinical interventions.
- Diagnosis: Understand how upset plots help in exploring the pattern between the overlapping positive microbiologic results within patient cohorts and supporting the decision making in clinical practice and clinical research.
- Adverse Event Analysis: Discover how upset plots facilitate adverse event monitoring and management by visualizing the co-occurrence of adverse events within patient populations.
- Missing Data Analysis: Gain insights into addressing missing data challenges using upset plots, identifying patterns and potential biases to ensure study validity.
- Medication Usage Profiling: Understand how upset plots elucidate medication usage patterns, concomitant drugs, and drug-drug interaction.
Throughout the tutorial, emphasis is placed on practical implementation using R programming (2). Detailed instructions and code snippets guide users through the process of creating impactful upset plots, empowering researchers and clinicians with the skills to effectively visualize and interpret complex data intersections.
Results: By mastering upset plots, stakeholders in clinical research and healthcare can enhance decision-making processes, improve diagnostic/treatment protocols, and drive advancements in evidence-based practice. This tutorial offers a comprehensive exploration of upset plots' applications, providing both theoretical insights and practical guidance. Through detailed case studies and hands-on demonstrations, researchers will develop a deep understanding of upset plots' potential to transform data visualization in clinical research and healthcare. With its focus on practical implementation and skill development, this tutorial equips analysts with the tools and knowledge needed to harness the full potential of upset plots, enabling them to tackle complex data analysis challenges with confidence and precision. From novice researchers to seasoned professionals, this tutorial caters to individuals at all levels, offering valuable insights and practical tips for leveraging upset plots to unlock hidden insights and drive data-driven decision-making in clinical research and healthcare.
Conclusions: Mastering upset plots is crucial for researchers and clinicians in unlocking hidden insights, driving data-driven decision-making, and advancing evidence-based practice in clinical research and healthcare. This tutorial offers valuable insights and practical tips for leveraging upset plots to transform data visualization and analysis.
All R codes will be shared will be shared.
References:
- Lex A, Gehlenborg N, Strobelt H, Vuillemot R, Pfister H. UpSet: Visualization of Intersecting Sets. IEEE Trans Vis Comput Graph. 2014 Dec;20(12):1983-92. doi: 10.1109/TVCG.2014.2346248. PMID: 26356912; PMCID: PMC4720993.
- Michal Krassowski. (2020). krassowski/complex-upset. Zenodo. http://doi.org/10.5281/zenodo.3700590
