Skip to content

Semi-Supervise cellular deconvolution of Bulk RnaSeq using NMF and CiberSortx, DCQ or others

Notifications You must be signed in to change notification settings

kalidouBA/DICEPRO

Repository files navigation

DICEPRO (Deconvolution with Iterative Completion for Estimating cellular Proportion from RNA-seq data)

Is a software package that has been meticulously designed to address the persistent challenges that cellular deconvolution methods have faced for years. One of its key features is its ability to overcome the limitations posed by incomplete reference matrix, which have hampered the accuracy and efficiency of many existing methodologies. In doing so, DICEPRO hopes to contribute to the advancement of RNA-seq analysis, offering a more complete and reliable solution for better elucidating the complexities of cell composition and gene expression.

DICEPRO’s core functionality lies in its iterative deconvolution process, which continues until convergence is reached. This approach involves joint updates of the reference matrix, which may contain missing cell populations, and the cell population abundance matrix, which contains newly discovered populations. Like other deconvolution methods, DICEPRO starts by fundamentally modelling the deconvolution problem. It relies on established techniques such as CiberSortX, DCQ or others for the supervised aspect, while using non-negative matrix factorization (NMF) decomposition for the unsupervised component. This hybrid approach enables DICEPRO to provide a robust and versatile solution for unravelling complex RNA-seq data.

Installation

BiocManager::install("kalidouBA/DICEPRO")

Running cibersortx to deduce the fractions of the different cell populations requires convigurations following the steps below:

Step 1:

Install Docker Desktop : (https://www.docker.com/products/docker-desktop)

Open Docker Desktop on your local computer and log in. Then, you open a terminal, and you type the following command: \> docker pull cibersortx/fractions

Step 2:

The next thing you need is a token that you will provide every time you run the CIBERSORTx executables.

You can obtain the token from the CIBERSORTx website: (https://cibersortx.stanford.edu/getoken.php).

Please note that each token is uniquely tied to your user account, and tokens are good for a specific time interval from date of request, so you will need to request a new token when an existing one has expired.

Step 3:

Once you have pulled the CIBERSORTx executable from Docker, and you have obtained a token from the CIBERSORTx website, you now have access to the CIBERSORTx Fractions executable and can run it following the instructions below.

Once the configuration is complete and docker is running, you can run DICEPRO with the token from the CIBERSORTx website cibersortx_email and cibersortx_token using the main function DICEPRO().
Other cell deconvolution methods are available to see with the help of the function running_method().