Genespector is a Dash/Plotly app for interactive visualisation of large scRNA-Seq datasets analysed using the scanpy/AnnData packages. Genespector allows responsive plotting of datasets up to 100,000 cells. It remains stable at least up to 250,000 cells but there is a visible delay in plotting.

Genespector is based on the Dash/Plotly framework: https://dash.plot.ly/

Python > 3.4 and pip are required. To install the package:

1. Clone the repository:

git clone https://github.com/Iwo-K/genespector

2. Install the dependencies

pip install -r ./genespector/requirements.txt

3. Install the package

pip install ./genespector/

A simple example of scRNA-Seq dataset displayed with Genespector can be found here: http://128.232.224.252/example/

For anyone interested in haematopoiesis, the Gottgens lab is hosting several Genespector apps displaying following scRNA-Seq datasets:

1. Mouse haematopoietic stem and progenitor cells:

1. http://128.232.224.252/sfdata/ - from the paper: Nestorowa et al. Blood 2016, PMID: 27365425
2. http://128.232.224.252/LKdata/ - from the paper: Dahlin et al. Blood 2018, PMID: 29588278

2. Human Bone Marrow cells (from Human Cell Atlas): http://128.232.224.252/HCABM/

Genespector simply requires an AnnData object to run. Most adata object should work straight away, but you may want to select features to be shown in the app by removing unnecessary information from the AnnData object. An example jupyter notebook with data preprocessing can be found here: https://nbviewer.jupyter.org/github/Iwo-K/genespector/blob/master/example_genespector.ipynb

The adata objects are imported as follows:

• Gene expression values are imported from the adata.X (or adata.raw.X, depending on the use_raw argument)
• Gene names are imported from the adata.var.index (or adata.raw.var.index, depending on the use_raw argument) and become available from the Color points menu.
• Numeric values present in the adata.obs slot are imported and become available from the Color points menu.
• Categorical data present in the adata.obs slot are imported and become available from the Color points menu. Plots are rendered in layers according to the selected value in Split data by menu. Note: only categorical columns are imported, columns containing strings are ignored. To change the column type use the method: .astype('category')
• All available coordinates are imported from the adata.obs slot and become available in the Choose the X and Y axis of the plot menu
• User-defined color scales for categorical data are imported from the adata.uns slot, as long as the names are matching. For instance adata.obs['louvain'] matches the entry adata.uns['louvain_colors'].

When dealing with large datasets, to save memory try using sparse matrices with normalised/log-transformed data and avoid using full matrices containing scaled data.

In python, where adata is the AnnData object of choice:

import genespector as gp
gp.make_app(adata)

Example output:

.........
.........
* Serving Flask app "app" (lazy loading)
.........
* Running on http://127.0.0.1:8050/ (Press CTRL+C to quit)

Copy the address to your browser or click the link.

make_app() accepts the following arguments:

• adata - an AnnData object
• use_raw - logical, whether .raw.X slot should be used instead of .X (default: False)
• make_var_unique - logical, whether .var.index should be converted to unique values using the var_names_make_unique() function from scanpy
• main_title and subtitle - string, titles displayed above the App
• layout - function, creating a dictionary which controls the website layout, needs to contain necessary elements for the app. For an example see the layout.py file
• server - logical or name of the server used for deploying the app
• url_base_pathname - string, specifies the url address for the app (default: '/')
• assets_folder - string, path to the folder containing static files, e.g. the .css file
• run - logical, whether the app should be run (if True) or return a dash.Dash (if False) instance of Flask app, useful for deploying the app on a server.

Interface is quite simple. The plot area allows zooming, selection of points and exporting to png files (in this cases removing axes may be useful).

Specific subsets (plotted as layers) of the data can be selected by clicking the legend (double-click to isolate a specific subset)

Subsets are specified from the dropdown menu. To colour-code by subsets select 'categorical data' from the 'Colour points' menu.

Gene expression values can be chosen from the Colour points menu.

The app can also be embedded within an existing website using <iframe> or integrated into another Flask application. In the latter case the arguments: server, url_base_pathname, assets_folder are passed into the dash.Dash() call. Setting run to FALSE will return the app object. For details see: https://dash.plot.ly/integrating-dash

1. Displaying multiple adata files (selection from a dropdown menu)
2. 3d coordinate system
3. Comparison of gene expression levels across clusters - violin plots