mixOmics

This repository contains the R package which is hosted on
Bioconductor
and our stable and development GitHub versions.

Installation

(macOS users only: Ensure you have installed
XQuartz first.)

From Bioconductor

The best way to install mixOmics is using Bioconductor. You can see
the landing page for the release version of mixOmics on Bioconductor
here.

Make sure you have the latest R version and the latest BiocManager
package installed following these
instructions.

## install BiocManager if not installed
if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

## install mixOmics
BiocManager::install('mixOmics')

## load mixOmics
library(mixOmics) 

From Github

Bioconductor versions are updated twice a year, between these updates
you can downlod the latest stable version of mixOmics from Github
using:

BiocManager::install('mixOmicsTeam/mixOmics')

You can also install the development
version for
new features yet to be widely tested:

BiocManager::install("mixOmicsTeam/mixOmics@development")

From Docker container

You can install our latest stable Github version of mixOmics via our
Docker container. You can do this by downloading and using the Docker
desktop application or via the command line as described below.

docker pull mixomicsteam/mixomics

docker images

$ docker images 
  > REPOSITORY                       TAG       IMAGE ID       CREATED         SIZE
  > mixomicsteam/mixomics            latest    e755393ac247   2 weeks ago     4.38GB

docker run -e PASSWORD=your_password --rm -p 8787:8787 mixomicsteam/mixomics

sudo docker ps

$ sudo docker ps
  > CONTAINER ID   IMAGE                   COMMAND   CREATED         STATUS         PORTS                    NAMES
  > f14b0bc28326   mixomicsteam/mixomics   "/init"   7 minutes ago   Up 7 minutes   0.0.0.0:8787->8787/tcp   compassionate_mestorf

docker stop f14b0bc28326

Contribution

We welcome community contributions concordant with our code of
conduct.
We strongly recommend adhering to Bioconductor’s coding
guide for
software consistency if you wish to contribute to mixOmics R codes.

Bug reports and pull requests

To report a bug (or offer a solution for a bug!) visit:
https://github.com/mixOmicsTeam/mixOmics/issues. We fully welcome and
appreciate well-formatted and detailed pull requests. Preferably with
tests on our datasets.

install.packages("renv", Ncpus=4)
install.packages("devtools", Ncpus=4)

# restore the renv environment
renv::restore()

# or to initialise renv
# renv::init(bioconductor = TRUE)

# update the renv environment if needed
# renv::snapshot()

# test installation
devtools::install()
devtools::test()

# complete package check (takes a while)
devtools::check()

Discussion forum

We wish to make our discussions transparent so please direct your
analysis questions to our discussion forum
https://mixomics-users.discourse.group. This forum is aimed to host
discussions on choices of multivariate analyses, as well as comments and
suggestions to improve the package. We hope to create an active
community of users, data analysts, developers and R programmers alike!
Thank you!

About the mixOmics team

mixOmics is collaborative project between Australia (Melbourne),
France (Toulouse), and Canada (Vancouver). The core team includes
Kim-Anh Lê Cao - https://lecao-lab.science.unimelb.edu.au (University
of Melbourne), Florian Rohart - http://florian.rohart.free.fr
(Toulouse) and Sébastien Déjean -
https://perso.math.univ-toulouse.fr/dejean/. We also have key
contributors, past (Benoît Gautier, François Bartolo) and present (Al
Abadi, University of Melbourne) and several collaborators including
Amrit Singh (University of British Columbia), Olivier Chapleur (IRSTEA,
Paris), Antoine Bodein (Universite de Laval) - it could be you too, if
you wish to be involved!.

The project started at the Institut de Mathématiques de Toulouse in
France, and has been fully implemented in Australia, at the University
of Queensland, Brisbane (2009 – 2016) and at the University of
Melbourne, Australia (from 2017). We focus on the development of
computational and statistical methods for biological data integration
and their implementation in mixOmics.

Why this toolkit?

mixOmics offers a wide range of novel multivariate methods for the
exploration and integration of biological datasets with a particular
focus on variable selection. Single ’omics analysis does not provide
enough information to give a deep understanding of a biological system,
but we can obtain a more holistic view of a system by combining multiple
’omics analyses. Our mixOmics R package proposes a whole range of
multivariate methods that we developed and validated on many biological
studies to gain more insight into ’omics biological studies.

Want to know more?

www.mixOmics.org (tutorials and resources)

Our latest bookdown vignette:
https://mixomicsteam.github.io/mixOmics-Vignette/

Different types of methods

We have developed 17 novel multivariate methods (the package includes 19
methods in total). The names are full of acronyms, but are represented
in this diagram. PLS stands for Projection to Latent Structures
(also called Partial Least Squares, but not our preferred nomenclature),
CCA for Canonical Correlation Analysis.

That’s it! Ready! Set! Go!

Thank you for using mixOmics!

What’s New

November 2024

• enhancement request
  #216
  implemented parallel processing using BPPARAM across all tune()
  functions
• feature request
  #335 added
  seed argument to perf() functions for better reproducibility
• feature request
  #334 added
  seed argument to tune() functions for better reproducibility
• bug fix implemented for
  #303 multiple
  solutions found in perf() returns error
• bug fix implemented for
  #307
  plotIndiv() ellipses colours not matching points, now sample group
  order is respected and colours can be customised for points and
  ellipses
• updated documentation to fix issue
  #297 broken
  link in bookdown vignette
• updated documentation to fix issue
  #296 typo in
  vignette

The performance assessment and parameter tuning workflow has been
streamlined as described in issue
#343

• New function: perf.assess() This function essentially runs perf()
  on final model but only returns performance metrics for the number of
  components used in the final model. Designed to be used in the final
  step of the workflow for quantifying final model performance. Outputs
  a list of values but no plotting functionality avaliable. See PR
  344 for more
  details.

• Additional functionality for tune() functions and new tune()
  functions created tune() can now be used in its original capacity
  (to tune number of variables and components simultaneously) or just to
  tune number of components by internally calling perf(). Designed to
  be used for tuning both components and variables to keep across
  (s)PCA, (s)PLS, (s)PLSDA, block (s)PLSDA and mint (s)PLSDA models See
  PR #348 for more
  details.

October 2024

** Version 6.30.0 **

Bioconductor release version
6.30.0
released end of October 2024 Minor bug fixes and updated deprecated code
and unit tests, no major code changes and no changes to user experience
of mixOmics.

• bug fix implemented for
  #293
  splsda() example code error

March 2022

• bug fix implemented for Issue
  #196. perf()
  can now handle features with a (s)pls which have near zero variance.
• bug fix implemented for Issue
  #192.
  predict() can now handle when the testing and training data have
  their columns in different orders.
• bug fix implemented for Issue
  #178. If the
  indY parameter is used in block.spls(), circosPlot() can now
  properly identify the $Y$ dataframe.
• bug fix implemented for Issue
  #172. perf()
  now returns values for the choice.ncomp component when nrepeat
  $< 3$ whereas before it would just return NAs.
• bug fix implemented for Issue
  #171. cim()
  now can take pca objects as input.
• bug fix implemented for Issue
  #161.
  tune.spca() can now handle NA values appropriately.
• bug fix implemented for Issue
  #150. Provided
  users with a specific error message for when plotArrow() is run on a
  (mint).(s)plsda object.
• bug fix implemented for Issue
  #122. Provided
  users with a specific error message for when a splsda object that
  has only one sample associated with a given class is passed to
  perf().
• bug fix implemented for Issue
  #120.
  plotLoadings() now returns the loading values for features from
  all dataframes rather than just the last one when operating on a
  (mint).(block).(s)plsda object.
• bug fix implemented for Issue
  #43. Homogenised
  the way in which tune.mint.splsda() and perf.mint.splsda()
  calculate balanced error rate (BER) as there was disparity between
  them. Also made the global BER a weighted average of BERs across each
  study.
• enhancement implemented for Issue
  #30/#34. The
  parameter verbose.call was added to most of the methods. This
  parameter allows users to access the specific values input into the
  call of a function from its output.
• bug fix implemented for Issue
  #24.
  background.predict() can now operate on mint.splsda objects and
  can be used as part of plotIndiv().

July 2021

• new function plotMarkers to visualise the selected features in block
  analyses (see https://github.com/mixOmicsTeam/mixOmics/issues/134)
• tune.spls now able to tune the selected variables on both X and
  Y. See ?tune.spls
• new function impute.nipals to impute missing values using the nipals
  algorithm
• new function tune.spca to tune the number of selected variables for
  pca components
• circosPlot now has methods for block.spls objects. It can now
  handle similar feature names across blocks. It is also much more
  customisable. See advanced arguments in ?circosPlot
• new biplot function for pca and pls objects. See
  ?mixOmics::biplot
• plotDiablo now takes col.per.group (see #119)

April 2020

• weighted consensus plots for DIABLO objects now consider per-component
  weights

March 2020

• plotIndiv now supports (weighted) consensus plots for block
  analyses. See the example in this
  issue
• plotIndiv(..., ind.names=FALSE) warning
  issue now fixed

January 2020

• perf.block.splsda now supports calculation of combined AUC
• block.splsda bug which could drop some classes with
  near.zero.variance=TRUE now fixed