A heatmap is a matrix of data points for a particular set of biomarkers, such as genes, at a particular point in time and/or for a particular tissue sample in the study, as measured for each subject in the study.
Data input requirements:
- Either one or two cohorts.
- At least one numerical HDD node with one or more biomarkers selected.
- Low dimensional numerical and categorical nodes are optional.
After fetching data the following control panel will be shown:
The panel provides the following options:
- Rows to show: change this number to control the number of rows to show in the final heatmap. The rows shown depend on the chosen Ranking Criteria.
- Group columns by: you can set this to either Node Order or Subject ID.
- Ranking criteria: choose the metric to apply biomarker ranking. This will determine the order of rows in the heatmap. Options include metrics based on Expression level, Expression variability, and Differential expression. The last option is only available when having defined two cohort subsets during cohort selection, see Heatmap: Differential expression.
The heatmap will appear after clicking Create plot.
By default the heatmap is sorted based on the chosen ranking criteria. The heatmap contains the following elements:
- Rows for each of the selected (or all) biomarkers for the selected data node. Clicking on gene identifiers takes you to external reference pages (GeneCards or EMBL EBI).
- Numerical or categorical nodes added will be shown on separate rows.
- Columns for each individual in the chosen dataset, with the identifiers as they are known in the tranSMART.
- Coloured squares based on the calculated z-score. The colour scheme can be changed in the Heatmap: Toolbar. Hovering your mouse over the squares provides additional information. By default green means a low z-score where red means a high z-score. This can be adjusted in the toolbar.
- Each row and column has a set of arrows that can be used to control the ordering of the heatmap. Small checkboxes allow users to highlight specific columns in the heatmap.
Below the heatmap itself you can find a table with detailed results for all computed statistics that are available in the Ranking Criteria section of the control panel.
Next to the Create Plot, Capture SVG button a Download button is available that downloads the input data data and the computed statistics.
The toolbar in the bottom right of the window provides a set of functionalities to change the current representation of the heatmap.
Marker statistic: a dropdown (default: coef) that allows choosing several statistics that can be used to display in the most left column of the heatmap. Available options: coef, variance, range, mean, and median.
Colour scheme: set the heatmap colours different multiple or single colour schemas, default is Red to Green Schema.
Zoom: make everything smaller or bigger.
Apply cutoff: remove rows from the heatmap based on a cut-off on the chosen ranking criteria. There is also a reset button.
Clustering: the toolbar allows the user to create clustering instead of normal ordering, using the R functions for
dist()for calculating distances and
hclust()(docs) for clustering. Computed are Euclidean and Manhattan distances with complete, average, and single clustering.
Based on the chosen clustering the order of columns and rows will change to reflect the computed clusters. Dendrograms are shown to display the results.
Clustering can be done for columns, rows or both.
Heatmap: Differential expression
When having defined two cohort subsets some of the aspects of the analysis will be different. For one, the summary page that is shown after Fetch data will show information for both subsets. The heatmap control panel will have the options for Differential expression enabled under Ranking criteria. This allows the users to order the rows based on one of multiple differential expression metrics.
The heatmap image itself will have an additional row to indicate to which subset an individual belongs. This bar allows researchers to easily identify the groups after performing ordering or clustering.
The table below the heatmap will show additional columns the additional options available in the Ranking Criteria section of the control panel (TTEST, LOGFOLD, PVAL, ADJPVAL, and BVAL). These measures that have been calculated between both subsets.