DEICODE is a tool box for running Robust Aitchison PCA on sparse compositional omics datasets, linking specific features to beta-diversity ordination.
UPDATE
All future development on Robust Aitchison PCA has been moved from this repository to Gemelli. The code in this respository will be maintained for the current functionality. Merging to Gemelli allows for the matrix and tensor based methods to be accessed in the same location, since they share some functionality.
Installation
To install the most up to date version of deicode, run the following command
# pip (only supported for QIIME2 >= 2018.8)
pip install deicode
# conda (only supported for QIIME2 >= 2019.1)
conda install -c conda-forge deicode Note: that deicode is not compatible with python 2, and is compatible with Python 3.4 or later. deicode is currently in alpha. We are actively developing it, and backward-incompatible interface changes may arise.
Using DEICODE inside QIIME 2
- The QIIME2 forum tutorial can be found here.
- The official plugin docs and tutorial can be found here.
- The in-repo tutorial can be found here.
Using DEICODE as a standalone tool
There are two commands within deicode. The first is rpca and the second is auto-rpca. The only difference is that auto-rpca automatically estimates the underlying-rank of the matrix and requires no input for the n_components parameter. In the rpca the n_components must be set explicitly. The structure of the commands follows the QIIME2 commands exactly and so questions about the use of the tool can be answered in the tutorial in the Using DEICODE inside QIIME 2 section above. However, an example analysis without the use of QIIME2 can be found here.
Using DEICODE as a Python API
The rpca functionality of DEICODE is also exposed as a python API. An example analysis without the use of the command line can be found here.
Other Resources
-
The code for OptSpace was translated to python from a MATLAB package maintained by Sewoong Oh (UIUC).
-
Transforms and PCoA : Scikit-bio
-
Data For Examples : Qiita