This plugin allows users to upload and context-specific -omics data files to analyze them in the context of acute inflammation and generate new information that can otherwise not be directly seen in the data itself.
First, the user uploads a data text-file containing fold change (FC) values of elements in one or multiple samples to the plugin (1) and select the column that contains the element (e,g, gene) identifiers (2) and the type of id (e.g. gene name, hgnc id, ...) (3).
The plugin is divided into two parts: The estimation of biological process (i.e. phenotypes) levels and the prediction of key regulators.
1. Phenotype Estimation
The AIR contains detailed information on the molecular pathways involved in the regulation of phenotypes. Because those pathways are directed, i.e. contain information on the type of regulation, positive or negative, we assume that we can predict changes in the phenotype activity based on the influence of their regulating elements and context-specific omics data mapped on the AIR. For instance, if positive regulators of a specific phenotype show an increase in their expression, i.e. have a positive logFC value, we assume that the activity of this particular phenotype has increased as well and vice versa. The algorithm, which predicts phenotype levels from the data, combines the FC value of elements with their network topology features to accurately assess their influence on the phenotype.
The „Phenotype Estimation“ plugin of the AIR integrates this algorithm, allows users to estimate phenotype levels, visualize the results directly on the AIR or download them for further analyses.
Before starting the analysis, the user can select how phenotype values will be normalized after the calculations (4). Because the estimated phenotype levels are non-measurable values and should therefore not be compared between phenotypes, we recommend normalizing each phenotype individually (between -1 and 1).
The calculations can also be performed using only absolute values (5), i.e. ignoring FC value signs and types of regulations. This might be useful when just investigating, how strong the phenotype is affected in each sample.
After the analysis is complete, the plugin offers various functions for displaying the results:
Overlays (7): The estimated phenotype levels can be converted into overlays and displayed on the map. There are buttons to hide and show all overlays simultaneously without having to toggle each individually in the Minerva overlay panel. The user can decide whether to add FC values from their data to the overlays (6) - however, this may significantly decrease performance when using huge datasets.
Image (8): One single image of the phenotype submap with overlays can be generated to get a quick overview of the most important processes at each step in the inflammation resolution timeline.
Table (9): All results are directly displayed in a table, which can be exported as a simple tab-separated text file. Clicking on a single value itself will pop up a new graph, containing information on the influence and FC values of all regulators. By clicking the checkbox in the first column, the values of the respective phenotype will also be displayed in a line graph below.
2. Key regulator prediction
Using network topology features, this tool can predict elements in our dataset that may be the most probable regulators for the observed changes (i.e. FC values) of the data sample. Highly probable regulators interact with elements in the dataset the same way as their fold change (e.g. a high negative fold change value should correspond to a close inhibitory regulation in the network).
To perform the analysis the user simply selects the sample (11) and, if desired, filters the target by their molecule type (e.g. transcription factor, miRNA, ...) (12). To enhance the accuracy of the prediction, (10) should be checked if the data file contains transcriptomics data. This way, the network topology feature for the calculations will be analyzed while considering transcriptional regulations between elements.
The predicted regulators are then displayed in a scatter plot (13) by their sensitivity (ability to regulate the elements in the data file as their FC values describe them) and specificity (inability to regulate elements with no or a zero FC value). The colors of the regulators are based on their type of regulation: positive, if the element induces the FCs, and negative if it induces the exact opposite changes. Elements marked as 'external' are not included on the maps and will link to their entry on public databases.