https://github.com/pachterlab/PCCA

Code for performing PCA followed by CCA
https://github.com/pachterlab/PCCA

Last synced: 28 days ago
JSON representation

Code for performing PCA followed by CCA

• Host: GitHub
• URL: https://github.com/pachterlab/PCCA
• Owner: pachterlab
• Created: 2018-07-03T00:26:32.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2018-12-02T17:34:47.000Z (over 5 years ago)
• Last Synced: 2024-05-22T02:14:08.918Z (about 2 months ago)
• Language: Python
• Homepage:
• Size: 142 KB
• Stars: 18
• Watchers: 6
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Lists

• awesome-multi-omics - PCA+CCA - Brown - [paper](https://doi.org/10.1371/journal.pgen.1007841) (Software packages and methods / Multi-omics correlation or factor analysis)

README

        
Principal Component Correlation Analysis (PCCA)

=======

This repository contains the analysis scripts necessary for reproducing the results in "Expression

 reflects population structure." It contains three main files.

- Snakefile: A snakemake file that can be used to reproduce the entire analysis from start to

 finish, incuding downloading the data from public locations.

- pca_cca/util.py: A utility module containing the methods used in the analysis - implementations

 of coupled PCA and CCA as well as the leave-one-out projections, regression comparison, and

 permutation tests.

- make_config.py: A utility for constructing the Snakemake config file, useful for adjusting

 the number of components used, and selecting the methods and populations to use in the analysis

Running `python make_config.py` will produce the Snakemake config file equivalent to the

standard analysis, the only change being the number of permutations is reduced from 10M used

 in the main analysis to 10K because this is by far the most time consuming analysis step.

 To replicate the analysis in the paper, simply clone and `cd` into the repository, then type

```

$: python make_config.py

$: snakemake results

```

Other targets include `projection_figures` which will make the PCCA projection and

leave-one-out cross-validation plots, `preprocess_data` which will do all steps up

to and including the quantification of corrected transcript levels and genotype PCs, and `get_data`

 which will simply download the necessary data from public locations.

# Dependencies

This was developed using Python 3.6.2 with snakemake 3.13.3, numpy 1.13.3, pandas 0.20.3

 sklearn 0.19.1 and seaborn 0.9.0. It also requires plink2 and kallisto be available on the

command line. To use peer for batch correction, peer must also be available on the command line.

We have included an environment.yaml file to assist with setting up an anaconda environment

that is sufficient.

```

:$ conda env create --file environment.yaml

:$ source activate PCCA

```

# Use as a library

This repository is currently indended for replication of the results in the manuscript. Advanced

users may be interested in using the utility methods in their own analysis, which is possible

by importing the utility module provided. At this time, the module is not documented for use

as a library in other analyses and such use is not officially supported. We intend to provide

a simple-to-use, well-documented implementation of our methods for use by others in their own

analyses at a later date.