LabRI Shiny Connect is the cloud-ready implementation of the LabRI Shiny Application for the estimation and verification of reference intervals in clinical laboratories. It preserves the full analytical structure of the LabRI method while adapting the application to a session-based, cloud-compatible mode: no figures, spreadsheets, intermediate files, .RData, .Rhistory, or HTML reports are written to persistent folders on the server. The application is designed for Posit Connect, Posit Cloud, and institutional Shiny servers, and a public demonstration instance is currently hosted on Posit Connect Cloud.
Repository contents:
app.R— Launches the Shiny interface and orchestrates execution of the LabRI method in a cloud-compatible environment.LabRI_script_connect.Rmd— Analytical script that implements the LabRI method and renders the HTML report during the Shiny session.www/— Static files used by the interface (logo, screenshots).
Deployment-specific files such as manifest.json or renv.lock may also be included depending on the target platform.
A public deployment is available on Posit Connect Cloud as a demonstration instance, allowing the LabRI method to be evaluated directly in the browser without installing R or RStudio.
The instance runs on the Posit Connect Cloud Free plan (4 GB memory, 1 CPU, 20 monthly active hours, max. 5 hosted apps). Under these constraints the demo may exhibit instability or memory-related failures with large datasets or reports containing many figures. It is intended strictly as a showcase, not as a production system, and should not be used to process sensitive laboratory data.
For institutional or production use, users are encouraged to fork the official repository — i.e., create a personal GitHub copy that can be freely modified, redeployed, and synchronized with the original:
🔗 Official repository: https://github.com/labrgrupo/LabRI_shiny_connect
From a fork, users can:
- Deploy on Posit Connect Cloud (paid tiers) with higher memory, CPU, and uptime.
- Deploy on a self-hosted Posit Connect server or institutional Shiny Server.
- Adapt
LabRI_script_connect.Rmdandapp.Rfor shinyapps.io, AWS, Azure, Google Cloud Run, Hugging Face Spaces, or Docker-based environments. - Customize the application for internal authentication, branding, or laboratory-specific workflows under the terms of the GPL-3.0 license.
The LabRI Shiny tool is distributed in two complementary implementations that share the same analytical method but differ in their output-management strategy:
- LabRI Shiny Application LabRI Shiny Application (this repository) — runs locally after R and RStudio are installed. Automatically creates
3_Outputs/Figures/,3_Outputs/Spreadsheets/, and4_Report_HTML/and writes figures, spreadsheets, intermediate files, and HTML reports to disk. Appropriate for single-user desktop workflows. - LabRI_shiny_connect — cloud-ready implementation for Posit Connect, Posit Cloud, or institutional Shiny servers. Operates in session-based mode, with no persistent files written to the server directory. Outputs are rendered inside the Shiny interface or in temporary session files, and the HTML report can be downloaded manually by the user.
A public demonstration of the cloud-ready version is available at labrgroup-labri.share.connect.posit.cloud (Posit Connect Cloud Free plan — intended as a showcase only, not for production use).
| Feature | LabRI Shiny Application (local execution) | LabRI_shiny_connect |
|---|---|---|
| Runs on the user's computer | Yes | Optional |
| Requires prior installation of R and RStudio | Yes | No (for end users in a deployed cloud environment) |
| Designed for Posit Connect / Posit Cloud | No | Yes |
| Automatically creates output folders | Yes | No |
| Automatically saves figures | Yes | No |
| Automatically saves spreadsheets | Yes | No |
| Automatically saves HTML reports to fixed folders | Yes | No |
Saves .RData / .Rhistory |
Possible | No |
| Uses temporary files only for session rendering | No | Yes |
| Displays report in the Shiny interface | Yes | Yes |
| Browser-based use without local execution | No | Yes |
The analytical method is identical in both implementations; only the output architecture differs. The local version is preferred for individual analytical work and for processing larger datasets without cloud memory constraints, while the cloud-ready version is preferred for institutional deployment, demonstrations, training environments, and multi-user access.
The initial interface captures the analyst's name in the Name of the Responsible Specialist section and accepts .csv, .xls, or .xlsx uploads in the Define the Dataset section, where the relevant data column is selected. A status bar tracks processing progress.
The LabRI Method is the analytical core of the LabRI System, implemented primarily in LabRI_script_connect.Rmd. It is organized in two modules:
- Estimation Module — adaptive, multi-criteria estimation of reference intervals via data cleaning, transformation, and clustering, using
refineR,reflimR, and the Expectation–Maximization (EM) algorithm (mclust,mixR). - Verification Module — three-level analysis assessing statistical uncertainty, equivalence, and concordance, ensuring the estimated intervals are reliable for clinical application.
This module provides an adaptive, multi-criteria approach for the indirect estimation of reference intervals, combining parametric and non-parametric percentile approaches according to the number of clusters in the truncated distribution.
- Multi-cluster distributions — The reference limits estimated by
refineRandreflimRare combined through the Centroid of Winsorized Reference Limits: the Two-stage Winsorization sub-algorithm first estimates robust winsorized limits, then the Hartigan–Wong Centroid Reference Limits sub-algorithm computes the centroid (x = lower limit, y = upper limit) for a centralized, stable estimate. When clusters are sufficiently separated, the EM algorithm further refines the result. - Single-cluster distributions — The EM algorithm applies parametric and non-parametric methods to derive the best reference interval estimate.
Multiple criteria and methods are combined to ensure robust and comprehensive estimation and verification.
Before routine clinical application — and especially when reference intervals are derived from indirect methods — laboratories must verify their reference intervals. The Verification Module performs a three-level analysis to assess whether compared reference limits are equivalent:
- First Level — Statistical Uncertainty. Evaluates the magnitude of statistical uncertainty associated with the reference limits. If uncertainty is within acceptable bounds, the analysis proceeds to the second level.
- Second Level — Distance Criterion Based on Equivalence Testing. Compares the LabRI-estimated reference limit with a comparative limit using equivalence testing to assess whether differences are large enough to be considered relevant from a practical or clinical standpoint.
- Third Level — Concordance Evaluation. Triggered when the second level suggests Possible Equivalence or Probable Equivalence. Incorporates confidence intervals and applies Fleiss' Kappa, Lin's Concordance Correlation Coefficient, and Flagging Rates to support robust verification.
The LabRI workflow is computationally demanding (reference interval estimation, mixture modeling, confidence intervals, plotting, and HTML rendering). In low-memory cloud environments — particularly free-tier services such as the public demo — the final HTML conversion step often requires more memory than the statistical calculations themselves. Large datasets or reports with many figures, interactive plots, or large HTML tables may therefore exceed the available memory.
If a memory failure occurs during report generation, options include:
- reducing the maximum subsample size;
- using a smaller dataset for demonstration purposes;
- running the Shiny application for local execution for larger analyses;
- forking the repository and redeploying on a Posit Connect environment with more RAM or on another cloud platform.
This typically reflects the memory ceiling of the hosting environment, not a defect in the dataset or in the LabRI method.
A usage tutorial covering the installation of R and RStudio and the operation of the Shiny tool is available on the Grupo Lab R website, which centralizes the documentation and supporting materials for the LabRI System and other tools developed by the group.
Submit suggestions and bugs at:
https://github.com/labrgrupo/LabRI_shiny_connect/issues
Email:
alancdias@hotmail.com · labrgrupo@gmail.com
Link to the publication:
Distributed under the GPL-3.0 license.