Jeff Barrett, Scott Russell, Janani Murthy
Aridhia Bioinformatics
Objectives: We propose to enhance the outreach and reliability of current open-science solutions supporting MIDD and precision dosing as part of a broader precision medicine strategy. Collaborative efforts to support expanded open-science solution functionality include the addition of nlmixr2 with the nlmixr team and Great Ormand Street Hospital and r-based system pharmacology capabilities in collaboration with ESQ Labs. The status and functionality of these solutions for R&D are described along with general guidance on the various mechanisms by which such solutions can be exploited in a Digital Research Environment (DRE) workspace.
Methods: The Data services component of the DRE provides meta-data catalogue, search, summary cohort analysis and orchestrated access request facilities in line with FAIR data principles. It allows data controllers to openly publish metadata about their datasets, control and approve access to those datasets and allow registered and approved researchers to query and build cohorts of interest. Once approved, researchers make use of a Workspace to work with data, build machine learning models, develop models, curate novel datasets and author collaborative documentation all with fully audited and traceable activity and code and data version control. Common components of a workspace would include: a shared file system for structured and unstructured data too allow workspaces members to work on a common set of data and code files; a relational database that can be used to contain original data or build novel datasets and provide SQL access from workspace tools; a Kubernetes cluster for hosting customized containerized applications; a secure token-based file upload mechanism to allow users to bring their own code and data into the workspace; an inbound ‘airlock’ to scan incoming files for viruses and provide a means for administrators to review incoming files prior to general access; R and Python coding environments to build apps and models; data exploration tools such as table sorting, filtering and pre-defined biomedical analysis modules; Git, to provide the ability to collaborate and version code and data within a workspace; and access to virtual machine compute, including GPU backed machines for large scale modelling purposes.
Results: The DRE supports the typical and iterative model development flow. Current nlmixr deployment will be demonstrated as part of the presentation. Reporting is supported within embedded tooling. PK Sim is currently launched in a Windows-based VM within a DRE workspace; operation proceeding exactly as with standard use. To further improve portability and usability of any open science application, consideration can be given to the use of a web interface and avoiding native operating systems interfaces such as windows or Linux. A containerized application with a web interface is portable, can be deployed on any operating system and provides the same experience to all users.
Conclusions: Containerized apps and R shiny allow for dedicated analysis-driven interfaces based application. It is recommended to use containerized apps when applications are expected to require high performance computing such as running the models against large datasets that consumes CPU time and memory. The benefit of running containers with a Kubernetes clusters is the parallel scale and ease of deployment. As open science solutions become more integrated into regulatory decision making and precision medicine strategies, the requirements for their qualification, validation and robust performance will increase. The opportunity to connect these solutions to a secure, collaborative data platform will enhance seamless interaction of a solution with data and benefit from audit trail tracking relevant interactions. We provide herein examples of open source tools along with a recipe for how to best interact with the DRE.
References:
[1] Wilkinson, M., Dumontier, M., Aalbersberg, I. et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3, 160018 (2016). https://doi.org/10.1038/sdata.2016.18
[2] Munafò, M., Nosek, B., Bishop, D. et al. A manifesto for reproducible science. Nat Hum Behav 1, 0021 (2017). https://doi.org/10.1038/s41562-016-0021
[3] Foster ED, Deardorff A. Open Science Framework (OSF). J Med Libr Assoc. 2017 Apr;105(2):203–6. doi: 10.5195/jmla.2017.88. PMCID: PMC5370619.
[4] Shaw DL. Is Open Science the Future of Drug Development? Yale J Biol Med. 2017 Mar 29;90(1):147-151. PMID: 28356902; PMCID: PMC5369032.
[5] Git, https://git-scm.com/, last accessed 3/9/2023.
[6] GitHub, https://github.com/, last accessed 3/9/2023.
Reference: PAGE 32 (2024) Abstr 10897 [www.page-meeting.org/?abstract=10897]
Poster: Methodology - New Modelling Approaches