Outliers and reference ranges in LB domain
A tool developed within the IHI VICT3R consortium to detect and flag unusual observations or extreme values in laboratory tests. Demo version limited to RAT and a selection of few strains. More selection criteria and several new features will be available in the next release.

What this application does

This application helps the user to make a complete outlier assessment. For each genotype (Specie/Strain) all available laboratory findigs will be shown and unusual values can be interpreted in a structured way, supported by statistical metrics.

Data preparation

The workflow starts with data subsetting, followed by the creation of two analytical datasets. One comprehensive of all available tests in any number of animals, suitable for univariate analysis, the other including only tests wich have been - all - observed for the same number of animals (to be tuned), suitable for multivariate analysis.

Outlier detection

The app supports several univariate approaches, allowing users to compare how different methods identify unusual observations.

Multivariate review

Mahalanobis distance and PCA extend the analysis by examining the joint structure of the selected variables and the profile of each subject.

Workflow overview

Step 1
Filter

Define the analytical subset. Currently only LB specimen and strain are supported

Step 2
Review

Review data completeness and set a threshold for the generation of the multivariate data set trying to balance the number of tests and number of animals.

Step 3
Detect

Run univariate outlier methods and compare their results.

Step 4
Explore

Use Mahalanobis distance, PCA, and subject profiles for multivariate interpretation.

Step 5
Export

Download summary tables and the final dataset with outlier information.

Dataset A: Univariate analysis

This dataset keeps the broader set of variables after filtering and can still include missing values. Its role is to support variable-by-variable outlier screening, where each measurement can be assessed individually.

Dataset B: Multivariate analysis

This dataset is more restrictive. Only variables that satisfy the completion threshold are retained, and only complete cases are used. This makes the dataset suitable for multivariate procedures.

Inside the analytical workspace

After entering the application, begin with data preparation, then continue to the outlier modules. Extreme measures will be detected by several approaches. Conconrdance matrixes are available to facilitate data interpretation.

Univariate section IQR, SD, percentile, Box-Cox, and reference interval methods.
Multivariate section Mahalanobis distance, PCA plots, and subject profile views.
Final outputs Summary tables, outlier results, and complete exportable datasets. Possibility to record the assessment in the database for future considerations.

Dataset Information & Creation

Two Datasets Will Be Created:

📊 Dataset for Univariate Analysis
  • All variables after categorical filters
  • Includes variables with missing data
  • Used for individual outlier detection
📈 Dataset for Multivariate Analysis
  • Only variables meeting completion threshold
  • Only complete cases (no missing values)
  • Used for Mahalanobis, PCA, etc.

Data Filtering

Step 2: Subset selection

What does this mean?
  • Variables with completion ≥ threshold → Used in multivariate analysis
  • Variables with completion < threshold → Only used in univariate analysis
  • Red line in plot shows current threshold

📊 Dataset for Univariate Outliers (Overall View)

Characteristics:
  • All subjects after categorical filtering
  • Includes all variables (even with missing data)
  • Used for detecting outliers in individual variables
Data Availability Pyramid (Comparison by Sex):

🔍 Filter Detailed Statistics

📋 Variable Statistics Details

Top Variables by Missing Rate:
Multivariate Dataset (B) Summary Statistics:

Outlier Detection Settings

Current Status:

                        

                      

                    

                    

                      

                        

                          
Boxplot Settings

                        

                        

                          

                            

                              
                            

                          

                          

                          
Outlier Detection Method:

                          

                        

                      

                    

                  

                  

                    

                      

                        

                          
Subject Outlier Summary

                          

                            
                          

                        

                        

                          

                          

                          
Outlier Matrix (First 10 variables):

                          

                        

                      

                    

                    

                      

                        

                          
Variable Boxplots

                          

                            
                          

                        

                        

                          

                            
                            

                              
                              
                            

                          

                          

                          

                        

                      

                    

                  

                

              

              

                

                  

                    

                      

                        
Comprehensive Outlier Analysis Settings

                      

                      

                        
Run All Outlier Detection Methods

                        
                        

                        
Methods Included:

                        
    
                          
  • IQR Method (1.5*IQR)
    • 
                          
  • Standard Deviation (3 SD)
    • 
                          
  • Percentile (1st-99th)
    • 
                          
  • Box-Cox
    • 
                          
  • Reference Interval
    • 
                        

                        

                        
                          
                          Download Summary Table
                        
                      

                    

                  

                  

                    

                      

                        
Outlier Detection Summary

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Visualization of Outlier Counts (Subjects Perspective)

                        

                          
                        

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Visualization of Outlier Counts (Variables Perspective)

                        

                          
                        

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Method Agreement Analysis

                        

                          
                        

                      

                      

                        
Consistency Across Methods:

                        

                      

                    

                  

                

              

              

                

                  

                    

                      

                        
Mahalanobis Distance Plot

                      

                      

                        

                          

                            

                          

                          

                            

                          

                        

                      

                    

                  

                  

                    

                      

                        
Outlier Controls

                      

                      

                        

                          
                          
                        

                        

                        
                        

                        
Outlier Summary:

                        

                        

                        
Download Results:

                        
                          
                          Download Outlier Results
                        
                      

                    

                  

                

                

                  

                    

                      

                        
Subject Profile (Normalized Parameters)

                      

                      

                        

                          

                            

                          

                          

                            

                          

                        

                        

                        
                          
                          Click on a point in the Mahalanobis plots above to view that subject's detailed profile across both sexes.
                        
                      

                    

                  

                

                

                  

                    

                      

                        
Detected Outliers

                      

                      

                        

                      

                    

                  

                

              

              

                

                  

                    

                      

                        
PCA Configuration & Variable Contributions

                      

                      

                        

                          

                            

                              
                              Settings
                            

                            

                              

                                

                                  

                                    
                                  

                                

                              

                              

                                

                                  

                                    
                                  

                                

                              

                            

                          

                          

                            

                              
                              Axis Selection
                            

                            

                              

                                

                                  
                                  
                                

                              

                              

                                

                                  
                                  
                                

                              

                            

                          

                          

                            

                              
                              Table Preview
                            

                            

                              
                              

                                
                                
                              

                            

                          

                        

                        

                          

                            Variable Loadings (Rotation Matrix)
                          

                          

                        

                      

                    

                  

                

                

                  

                    

                      

                        
PCA Biplot - Male

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
PCA Biplot - Female

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Variance Explained (Scree Plot)

                      

                      

                        

                      

                    

                  

                

              

              

                

                  

                    

                      

                        
Complete Dataset with Outlier Flags

                      

                      

                        

                        

                        
                          
                          Download Complete Dataset
                        
                      

                    

                  

                

              

              

                

                  

                    

                      

                        
Outliers by Study - Male

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Outliers by Study - Female

                      

                      

                        

                      

                    

                  

                

                

                  

                    

                      

                        
Detailed Study Statistics