Integrative data analysis approaches have been successfully used to analyze multiple datasets simultaneously in order to compare the results of independent experiments. This approach can also be used to investigate which features (e.g. genes or pathways) are consistently down or up regulated in different studies, or across different conditions of the same study.
To perform this kind of integrative analysis, we use an approach called rank-based meta-analysis, to identify consistently upregulated features across studies or conditions. Meta-analysis provides a variety of benefits, including the modeling of the inter-study variation, the identification of novel differentially regulated features, and the discovery of biomarkers in disease conditions.
The UniApp performs the meta-analysis by combining the results obtained during differential analysis, geneset enrichment analysis, variation analysis, text mining, patent mining (Section [link]), cluster marker genes analysis, gene association analysis or results of other rank-based meta analyses. It is imperative that there is a sufficient overlap between features in analyses included, either for genes or pathways. It is not possible to combine results examining genes and pathways in the same meta-analysis.
1 Algorithm settings
1.1 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 to a section where the analysis of interest (in this cased Rank-based meta analysis) can be selected.
In order to ensure efficient organisation, a name and description must be assigned to the analysis under the appropriate fields. Subsequently under "Choose algorithm to run your analysis" Rank-based meta-analysis must be selected.
1.2 Selecting data
In the field "Choose track element", input analyses can be selected. They are listed in the field "Selected analysis results".
1.3 Setting parameters
Once all input analyses have been selected, algorithm parameters can be specified. Ranks obtained from every input analysis can be inverted if downregulated features are to rank higher (as mentioned above, by default the upregulated features are ranking higher).
If an input analysis is of type cluster marker genes or gene association analysis, specific clusters can be selected to be included in the meta-analysis under the field "Select columns". Afterwards, individual selected clusters can be inverted if so desired.
1.4 Running the analysis
Once all parameters have been set, the analysis can be executed by clicking the button "Run" on the top-right of the screen.
2 Accessing results of the analysis
You will be redirected back to the Tracks page where new placeholder ("Rank-based meta analysis") will appear. Click "Select" to view the newly executed analysis.
The selected analysis will appear at the top of the list and a new option "View Interactivelly" will be available. Press on "View Interactivelly" and wait for results to load. You will be redirected to the interactive plot page.
3 Interactive plot page
3.1 Ranked-based meta-analysis results
The output of this analysis is a table including a list of overlapping features (e.g. gene names or pathway names) and:
- rankings for every feature in each input analysis, or
- ranking for every feature aggregated across all input analyses based of meta-score.
By default, the higher the meta-score, the more a feature is upregulated (or more suited in text mining and patent mining), e.g. meta-score of 95 means that the feature is in the top 5% of upregulated features.
For each input analysis, features are first ranked based on results, yielding a ranked lists. From these ranked lists, meta-scores for each analysis are computed (the highest ranking feature gets a meta-score of 100, the lowest ranking 0). Afterwards, meta-scores for every feature are aggregated across all analyses by calculating their product. These aggregated meta-scores are then ranked again and new meta-scores are computed, resulting in a single meta-score for every feature.
3.2 Select input
In Select input tab you can choose which features to plot and which datasets to include or exclude from the meta-anaylsis via the multi-select menus.
- In Features to plot, a feature of interest must be selected in order to generate the plot. The selected gene will be carried over to all other plots of the meta-analysis.
- In Datasets to plot, you can choose which of the initial input datasets to plot. These datasets will also be used in recalculating the meta-analysis score.
From the Feature to plot menu all of the features that are detected in all of the input datasets are displayed. It is possible for a feature to be detected only in a subset of all input datasets. If you select such a feature then only the datasets in which that feature is detected will be available in the Datasets to plot menu. If you previously made a selection of datasets and have later selected a feature that is not detected in one of the datasets, that dataset will be automatically dropped from the selection.
Once the input is changed a new plot will be rendered and the meta-score recalculated once the Update plot button is clicked. The input selection will be carried over to all other plots and table of the meta-analysis.
3.3 Plot types
The Rank-based meta analysis currentlysupports five types of visualizations: Dot plot, Spider plot, Bar plot, Violin plot, String plot and Table. Examples below in the respective order.
3.3.1 Dot plot
The dot plot shows the ranking of the selected gene in the individual selected datasets.
3.3.2 Spider plot
The spider plot shows the ranking of the selected gene in the individual selected datasets.
3.3.3 Bar plot
The bar plot shows the ranking of the selected gene in the individual selected datasets.
3.3.4 Violin plot
The violin plot is used to examine where the selected gene is located in a distribution of genes based on log2fold changes. The violin plot can only be visualized if only the results of diffirential expression analyses are selected.
In case you have selected any other type of dataset except the differential analysis while trying to visualize a violin plot the following message will be displayed: "To visualize a violin plot please only select differential analysis results". Once the dataset selection is made the violin plot will render.
3.3.5 String plot
The String plot shows shows the ranking for the selected feature aggregated across all input analyses based on meta-score. You can visualize more than one feature on multiple strings that have the same horizontal axis by selecting more than one feature in the Feature to plot menu:
Note that the options available in the Datasets to plot menu are dependent on the choice made in the Feature to plot menu. In the Feature to plot menu all detected feautres in all of the input datasets are displayed. When you select one or more features in the Feature to plot menu then only the datasets in which these features are detected will be available in the Datasets to plot menu. If you previously made a selection of datasets and have later selected a feature that is not detected in one of the datasets, that dataset will be automatically dropped from the selection.
Finally with the table view it is possible to view tabular results and statistics on the rank-based meta-analysis that was perfomed with the current input selection.
Was this article helpful?
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
We appreciate your effort and will try to fix the article