# Documentation

The use of CZViz is normally quite intuitive, and the user in a hurry should be able to skip the documentation without encountering any problem. However, for more specific information, you can refer to the documentation below.

## App initialization

The app automatically initializes itself when loading the corresponding Home webpage of this website. If the app is stuck on for more than a couple of seconds "Loading...", a page refresh should solve the problem.

## Gene selection

You can choose your gene on interest in the panel on the right. Three alternative representations are then proposed: either the spatial scatterplot profile of the gene, represented for all the timepoints and all replicates, either the temporal scatterplot profile of the gene, represented for each layer, either both at the same time in a 3D graph. For each of these representations, the average among the replicates, for a given time or space condition, is also given as a line.

Finally, a static table representing the gene data appears at the bottom right of the page.

## Statistical analysis

The tab Statistical testing enables for applying different types of linear regression on the data using ordinary least squares (see the Analysis page). You can choose the desired type of analysis using the dropdown menu on the right.

For the spatial regression, a representation of the spatial profile of the gene, including all replicates, is given for each time condition. The model prediction is plotted as a blue line. If the retained model is not flat, the the parameters kept associated to the zonation are given.

For the temporal regression, three representations are proposed. First, a sum up of the phase and amplitude of the gene for each layer is given in a polar chart. Similarly, the average temporal profile across each layer is also given. Finally, a representation of the temporal profile of the gene is given, including all replicates, with the model prediction as an orange line. If the retained model is not flat, the p-value associated to the rhythmicity is given, as well with the parameters kept.

For the spatiotemporal regression, one or two representation(s) is (are) proposed, depending if the retained model has rhythmic parameters evolving with the layers. The global fit is always given, as a 2D surface going through the experimental points. If the model has evolving rhythmic parameters (that is, if the phase of the signal is defined and variable, i.e. $$\exists i \ne 0 | (a_i, b_i) \ne (0,0)$$ ), a polar chart representing the fit of the conserved model (as a curve) against the individual temporal fits (as markers) is given. If the retained model is not flat, the p-value associated to the model is given, as well with the parameters kept.

## Validation

In order to assess the validity of the data, it can be useful to compare the loaded dataset to a previously existing dataset. The tool implemented in this tab offers the possibility to make profile comparisons for all the common genes of the current dataset and the dataset coming from Atger and al. It is accessible via the dropdown menu of the tab Statistical testing.

Since the dataset from the current study has a supplementary spatial dimension, gene profiles are not directly comparable. Moreoever, pre-formatting of the two datasets is the same, leading to different units and range of expression. Two steps are thus involved, in order to obtain comparable data.

• The different layers are averaged using a a weighted arithmetic mean. Weights are optimized in order to maximize the global correlation between the gene profiles of the two dataset.
• Both datasets are standardized : for each gene profile, the mean is removed, and the resulting datapoints are divided by the standard deviation of the profile.

The result is represented using both the standardized temporal profiles, as well as a polar chart representing the phase and amplitude obtained from each dataset using a sinusoidal linear regression.