Fitnoise

Fitnoise is a Python 2 library for statistical analysis of RNA-Seq, PAT-Seq, and microarray data using linear models.

An R wrapper is provided to allow access from R.

Fitnoise uses the Theano deep-learning library for speed.

Fitnoise is developed by Dr. Paul Harrison for the RNA Systems Biology Laboratory, Monash University.

Overview:

• Poster presented at Lorne Genome Conference 2015

  (An earlier poster presented at ABiC 2014 describes the previous R based Fitnoise.)

Documentation:

• What is Fitnoise?
• How to use Fitnoise
• Assessing the quality of a fit
• Noise models available
• Control genes (and what to do if you don't have replicates)

Download:

• Download latest or older versions

Links:

• Fitnoise on Github
• Fitnoise on PyPI

Installing into a virtualenv from source

This is the easiest way to try out Fitnoise.

The following creates a virtualenv in directory venv for both Python and R:

./make_virtualenv.sh venv

To freshen the virtualenv after pulling a new version of Fitnoise from github or hacking on the code:

./freshen.sh venv

Installing globally

On the python side, Fitnoise requires Theano. On the R side, Fitnoise requires rPython, and jsonlite, and limma.

Installing dependencies:

apt-get install python-pip python-numpy python-scipy r-base
# (or whichever package manager is appropriate to your Linux distribution)
# (MacOS users can use brew and Anaconda Python)

pip install --upgrade git+git://github.com/Theano/Theano.git

R
  install.packages(c("rPython", "jsonlite"))
  source("http://bioconductor.org/biocLite.R")
  biocLite("limma")

To install Fitnoise with pip:

pip install --upgrade fitnoise

python -m fitnoise
# This prints out instructions to install the R component

Alternatively, to install Fitnoise from source:

python setup.py install
R CMD INSTALL fitnoise

Alternatively, to install the development version of Fitnoise directly from github:

pip install --upgrade 'git+https://github.com/pfh/fitnoise.git#egg=fitnoise'

R
  install.packages("devtools")
  devtools::install_github("pfh/fitnoise", subdir="fitnoise")

References

Fitnoise re-implements various features of the limma Bioconductor package:

http://bioinf.wehi.edu.au/limma/