Personalizing drug doses in the cloud using R
R has a long history in PK/PD modeling, and it has been heavily used both in research and clinical practice as well, but making these R packages available outside of the R community has its (technical, compliance and UX) challenges that even hosted Shiny apps cannot easily solve yet.
Rx Studio is a clinical decision support webapp and cloud platform for precision drug dosing, using peer-reviewed and custom-built PK/PD/PGx models implemented in R (containerized Plumber APIs), hosted in a HIPAA-compliant infrastructure (AWS and GCP services), and made available to end-users via a user-friendly and configurable web interface (Angular and Nebular).
This talk will provide an introduction to the importance of personalizing drug doses (especially in critical care), followed by a short demonstration of the system, then focusing on the applied PK/PD models and the Markov Chain Monte Carlo (MCMC) simulations using adaptive Metropolis for finding the Bayesian PK parameters for a specific patient, and providing a high-level overview on the cloud infrastructure as well.
No math, stats, or prior medical knowledge is required; this is more of an ML use-case in healthcare.
Gergely Daróczi is an enthusiast R user and package developer; Ph.D. in social sciences; former assistant professor in Sociology, currently a lecturer at the Business Analytics program of CEU; 15+ years of industry experience in data science, engineering, cloud infrastructure, and data operations at SaaS, fintech, adtech, and healthtech startups with a strong interest in building scalable data platforms on the top of R and AWS. He maintains a dozen CRAN packages related to using R in production (automated reports, logging, database connections, API integrations), co-authored several journal articles in social and medical sciences, and wrote a book on “Mastering Data Analysis with R”.