What this app does:

This is a web app containing exploratory graphs for spatialomics data. It has been designed while working with spatially resolved proteomics and transcriptomic data from the GeoMx DSP and will draw 5 interactive plots on any dataset that you give it. The plots are:

A parallel categories plot of the selected columns
A scatter plot of the data (with spearman's rho and p-value)
3D UMAP
Volcano plot of the data (with univariate results of a Linear Mixed Model)
And a univariate Cox regression horizontal bar plot

Notes on data input

While the app is robust in reading file formats, it assumes that the uploaded data is formatted so that all information is stored in columns. All biomarker counts need to be in columns and all factors also need to be in columns and one column needs to be some kind of identification to track the source (patient).

Quick start:

The main script is app/mk_dash.py but before running it you need to setup an environment, this can be done using Conda.

Windows users:

After installing find the anaconda prompt
Use anaconda prompt to run the specified commands!

Linux and mac users:

You probably know how to use conda and a terminal! :)

(if you don't: look for the link to conda docs above!)

Open a terminal in path/to/mk_dsp_explorer and run:

conda env create -n dashenv -f env/environment.yml

Next, activate your new environment:

conda activate dashenv

In the future you may wish to list your environments: conda env list

Configuration

All settings made in the app are now also written to conf.json, this file can be backed up when switching.

When you are ready you can run the main app by doing:

python app/mk_dash.py

NOTES:

Expected data types of the different analyses:

Parallel plot - categorical data will be transformed to numbers ordered by sorting otherwise colors won't work!
LMM - binary or ordinal encoding of the fixed effect, continuous values for biomarkers
Cox - binary, ordinal or continuous - estimates a ratio that is influenced by the order of values
Correlation - continuous (the idea is to check for correlation of various biomarkers)
3D UMAP - distance is measured for the biomarkers, which are logscale continuous