- Clicking in the picture on each of the highlighted areas will bring up more information on it.
- The chart and the axes can be zoomed as described in chart controls.
- When samples and/or probes are selected with the mouse, the selections in the plate view, plate list view, sample view and probe view are synchronized.

The heatmap is a grid of cells. Each column corresponds to one probe. Each row corresponds to one sample. The cell is colored by the fold-change of the probe in this sample relative to the average of all samples. The level in a particular sample is based on an average of all the particles in the sample, as described under statistics. The value is potentially blank-subtracted, negative-subtracted and normalized, as described under data
processing. The color scale bar at the upper right shows the fold change.

Hovering over a rectangle brings up detailed information for the sample/probe combination. The arithmetic and logarithmic values are presented. Also the data values from raw through blank subtraction through negative subtraction through normalization are shown.

The probe names are listed along the top of the axis. If there are too many names, some names will disappear to allow the remaining names to be displayed.

The probes are clustered as described under clustering. Probes that have similar expression levels to each other are placed closed to each other. The shorter the branches, and the fewer the branches, joining two probes, the closer they are to each other in expression pattern across samples.

The sample names are listed along the top of the axis. If there are too many names, some names will disappear to allow the remaining names to be displayed.

The samples are clustered as described under clustering. Samples that show similar probe expression patterns are placed close together. The shorter the branches, and the fewer the branches, joining two samples, the closer they are to each other in expression pattern. Replicates will generally cluster closely together.

The color scale is scaled so the darkest color corresponds to the lowest numerical value in the array, and the brightest color corresponds to the highest numerical value in the array. Average-expression numbers are white. The numerical scale is plotted along the color scale for reference. A negative sample will appear as a dark horizontal bar. A blank probe will appears as a dark vertical bar.

The different variables that have been defined appear. They are also colored according to their values. For real-valued variables, such as the sample quality ("qc"), the colors run from the lowest to highest value of the variable. For discrete variables, such as patient gender, or group ID, the color corresponds to the alphabetical order of the group name.

- The transformation drop-down menu allows selection between linear data, log data, and fold-change data.

- With linear data, a single large value can result in a binary heatmap -- one orange and everything else blue.

- Log data spreads the values through color space in a more interesting way
- Fold-change data shows the log10 of the ratio between a probe value to the average of all the other samples for that probe (i.e. relative to the column average). Fold-change is often the best way to view a heatmap because under-expressed and over-expressed probes are given equal prominence.

- After blank and negative subtraction, it is possible for a probe to be negative in a sample. To avoid taking the log of a negative value, a floor of 1.0 is applied.
- If there is one probe with a very high range of over and under expression, it may dominate the heatmap. To visualize the more moderate probes, the color range can be zoomed using the sub-range controls.
- Inverting the color map swaps the ends.

- Analysis - the axis can be sorted as
- None - probe order
- Cluster - order defined by the clustering tree
- Group -- the probes are ordered by most significantly differentially expressed.

- Metric - different metrics can be used to estimate the "similarity" between each pair of probes.

- RMS compares two probes as the root mean square difference of the probe levels
- Pearson compares two probes as 1-√ R
^{2}where R^{2}is the Pearson correlation coefficient of the probes. - Since probes are often of different orders of magnitude, the
**Pearson**comparison is more likely to be useful for comparing one probe to another, since it was developed in the context of correlating variable Y to variable X, which may not even be measured in the same units

- The branch law affects the drawing of the tree branches. If the branches are of very different lengths, some of the horizontal tree segments may be impossible to distinguish. A branch law of 1.0 draws branches in linear proportion to the clustered branch length. A branch law of 0.5 draw branches as the square root of the clustered branch length. A branch law of 0.0 draws branches independent of computed length, so that the length of the branch reflects only the depth in the tree.
- The checkbox Show Labels turns the probe names on and off
- The checkbox Draw Tree turns the drawn tree on and off. The order is preserved even if the tree is not visible.

- Analysis - the axis can be clustered, or not clustered
- Metric - different metrics can be used to estimate the "similarity" between each pair of samples.

- RMS compares two samples as the root mean square difference of the sample levels across all the probes.

- Pearson compares two samples as 1-√ R
^{2}where R^{2}is the Pearson correlation coefficient of the samples across all the probes. - Since sample values tend to be of like magnitude, either Pearson or RMS metrics are likely to result in similar results.

- The branch law affects the drawing of the tree branches. If the branches are of very different lengths, some of the horizontal tree segments may be impossible to distinguish. A branch law of 1.0 draws branches in linear proportion to the clustered branch length. A branch law of 0.5 draw branches as the square root of the clustered branch length. A branch law of 0.0 draws branches independent of computed length, so that the length of the branch reflects only the depth in the tree.
- The checkbox Show Labels turns the sample names on and off
- The checkbox Draw Tree turns the drawn tree on and off. The order is preserved even if the tree is not visible.