Awesome

NMFreg tutorial

Did you ever want to try NMFreg on your data? Here is the tutorial!

Coming soon! Examples of other applications :)

Do you have an application where NMFreg might help deconvolve your composite measurements aided by a labeled reference? Send me an email!

How do I run this?

There are two options:

Locally

Note: This requires standard scientific Python 3 environment. A simple way of getting that is installing Anaconda.

Run the following commands in your terminal:

git clone https://github.com/tudaga/NMFreg_tutorial
cd NMFreg_tutorial
jupyter notebook NMFreg_Tutorial_cerebellum_puck180430_6.ipynb

Remotely via Google Colab

Click on <a href="https://colab.research.google.com/github/tudaga/NMFreg_tutorial/blob/master/NMFreg_Tutorial_cerebellum_puck180430_6.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>.

Intro

The notebook NMFreg_Tutorial_cerebellum_puck180430_6.ipynb goes over a cerebellum example. The basic steps are:

Run NMF on a labeled single-cell RNA-seq cerebellum dataset to derive an interpretable basis.
Regress the Slide-seq beads onto the basis via NNLS to deconvolve each bead into proportional contributins from each cell type.
Bonus Get a heuristic measure on the certainty that a bead contains mRNA from a single celltype.

If you want to learn more about NMF, watch my lecture on it here.

Reference

This work is featured in the flagship paper for Slide-seq: A scalable technology for measuring genome-wide expression at high spatial resolution.