This document has last been compiled on 2021-12-17 01:10:41.

Note that what follows is generic text, repeated in many files, to describe the different heatmaps that can follow. What is actually shown in this document may only be a subset of these heatmaps.

We show side-by-side heatmaps if there are sufficient number of genes.

There are 4 possible heatmaps that can be shown for any set of genes. For each heatmap, the genes are on the rows, and the samples are on the columns, and each entry corresponds to the value of expression for that sample and gene. The samples grouped by genotype and their watering condition (i.e. treatment), and have color markings to indicate the sample’s condition/genotype/TP, as well as the TP in which the plants in the block actually flowered. The legend below this text gives the colors that will annotate the samples in future heatmaps (and the heatmaps will not have this legend because it takes up too much space).

In the multiple heatmaps, the genes are clustered based on the values in the first heatmap (starting in the upper left), and subsequent heatmaps are in the same order as the first heatmap. The heatmaps differ in how the expression data is shown. The colors used are generally the same or similar, but the the interpretation and scale are quite different. Here are the possible options, and each heatmap is clearly labeled in the title of the heatmap as to which it is:

  1. “Centered” This means the mean value of each gene is subtracted, so that the absolute level of the expression is not shown but rather the expression of the gene relative to other samples value of this gene. The centering is done across all the samples shown. This means if there is a difference in the absoulte level of expression between genotypes, then one genotype will be all low (blue) and the other genotype is all high (low). This doesn’t mean that there is no difference between the TPs, but just that the difference between the genotypes dominates. These types of samples will invariably cluster together, but should not be over interpretated (see genotype analysis document for more look at genotype differences).

  2. “log-Ratio” This means that the value shown for each gene/sample combination is the ratio of the log-expression of the treatment to the control, with plants from the same replicate and condition being paired together for the ratio. (This also means that there is no “control” sample colors shown in the top).

  3. “Centered Separately” This means that the centering described above was done separately by genotype. In otherwords, the mean was calculated per genotype, so that the values are only the relative expression within a genotype. This allows for being able to see the patterns across the TP of the gene, even if there is a different level of expression, but hides differences in a genotype.

  4. “Absolute” This means that the absolute value will be shown, with no centering. This is often not terribly informative, except in the case when a gene has extreme differences in expression between the genotypes, such as off and on (small differences will not necessarily be distinguishable on the scale of the heatmap).

A note on the color scale Except for “log-ratio” heatmaps, the color scales for the heatmap are data-dependent, meaning bright yellow might be a different value from one set of genes to another. Furthermore, the color scale legend might look strange because much of the scale is either all yellow or all bright blue (the extreme colors). This is because we create evenly spaced breaks up to the 99% of the data. This means, very large data values all get the same value – you can’t really tell the difference between them. This is because otherwise a single outlying point can steal the entire color spectrum, leaving the rest of the data all of the same color.

An exception to this is the log-ratio, where we set fixed scales across all of the heatmaps. This means that a log-ratio between -.25 and .25 (ratio less than 1.28) shows up as white, and then after that it is evenly spaced on the log scale between .5 and 4 (and equivalently for inverse); all log-ratio values greater than 4 are equivalent for the color scale.