TABLE OF CONTENTS

• Introduction
• 1 Creating a plot
• 2 Selecting data
• 3 Setting parameters
  • 3.1 Differential analysis design
  • 3.2 Differential analysis algorithm
  • Sets the type of model to use, can be Limma or MAST. They are both generalized linear models (see Section TODO for more information). Limma is a popular method to analyze bulk datasets, while MAST is more accurate for zero-inflated single cell data. TODO (explain T-test and Wilcoxon)
• 4. Performing the differential gene analysis
• When the parameters are all set-up, you can click on the Run button to compute the differential gene analysis. 
• 5 Differential gene analysis interactive plot page
  • 5.1 Plot parameters
  • 5.2 Highlight features
    • 5.2.1 Highlight features - gene metrics
    • 5.2.2 Highlight features - custom gene list
  • 5.3 Marker format and color for numerical variables
  • 5.4 Details 
    • 5.4.1 Grid style
    • 5.4.2 Title style
    • 5.4.3 Plot margins
  • 5.5 Axes style
  • 5.6 Export settings
• 6 Differential analysis video tutorial
• 7 Useful links

Introduction

    Differential analysis is one of the most commonly performed analyses when interpreting omics data. Differential analysis means taking the (normalized) data and performing statistical analysis to discover quantitative changes in expression/abundance levels between experimental groups. For example, for each feature in the data we can perform statistical testing to decide whether an observed difference (change) in expression/abundance is significant or not, which means checking whether the change is greater than what would be expected due to natural random variation. Basically, differential analysis is used to determine whether there are any features that are significantly different between two groups.

    The UniApp uses linear models (limma, MAST, T-test and Wilcoxon). This statistical technique is highly flexible and can accommodate a large variety of experimental designs, correct for confounding variables, etc. The results are summarized through a volcano plot, but the full results can still be explored through an interactive table.

1 Creating a plot

As a first step of the analysis, a plot must be created by clicking on the create plot icon. This will lead you to a section where the analysis of interest can be selected.

In order to ensure efficient organization, a name and description must be assigned to the analysis under the appropriate fields. Subsequently, select "Variation Analysis" under "Choose algorithm to run your analysis". Press "Select Algorithm" which will lead to the second column opening on the left.

2 Selecting data

Now the required project, track and plot can be selected. Differential gene analysis is done based on the results of Data Pretreatment, make sure to select the right track element. Press "Select Data" when the required fields are filled out and the third column will open on the left.

3 Setting parameters

• Design: here you can define between which groups you wish to perform the differential analysis. Groups are defined by metadata categories.
• Algorithm: the type of model to use, can be Limma or MAST. They are both generalized linear models. Limma is a popular method to analyze both bulk and single cell datasets, while MAST is more accurate for zero-inflated single cell data.

3.1 Differential analysis design

In the Design tab you can choose between which metadata groups you wish to perform the differential analysis. 

• Observation:  set metadata variable whose groups you wish to compare. Note that this is supported only for categorical variables.
• Reference group: sets reference group of the selected metadata variable. This can be usually be the control or healthy group of your experiment. Genes expressed highly in the reference group will appear as downregulated and with a negative logFC. You can select multiple groups.
• Experimenal group: sets experimental group of the selected metadata variable. This can be usually be the treatment or diseased group of your experiment. Genes expressed highly in the experimental group will appear as upregulated and with a positive logFC. You can select multiple groups. 
• Covariates: the covariates to include during the modelling. The covariates can be chosen from the variables present in the metadata.
• Scaling: In the case of metabolomics and proteomics data you will see this additional option under scaling (the following options are available: None, Auto, Center, Scale, Range, Pareto, Vast, Level). 

There must be no overlap of observations/groups between the reference and experimental groups. Additionally, you cannot select as a covariate the same group you are using to perform the comparison.

3.2 Differential analysis algorithm