bioconductor-mixomics

name: bioconductor-mixomicsdescription: Multivariate methods are well suited to large omics data sets where the number of variables (e.g. genes, proteins, metabolites) is much larger than the number of samples (patients, cells, mice). They have the appealing properties of reducinwhen_to_use: "Use when: Performing unsupervised exploratory analysis and dimensionality reduction using pca and spca.; Integrating multiple omics datasets measured on the same samples (N-integration) using DIABLO.; Conducting supervised classification and feature selection for multi-class problems using splsda.. Not for: For simple univariate differential expression, use limma or DESeq2 instead because mixOmics is designed for multivariate feature selection and integration.; For raw mass spectrometry or sequencing read preprocessing, use xcms or edgeR for normalizati"user-invocable: false

mixOmics

Dependencies & Environment

Package-intrinsic requirements from the Bioconductor landing page — reproduce in any R environment.

• Version: 6.36.0 · Bioconductor: 3.23 · R: ≥ 4.6
• Depends: MASS, lattice, ggplot2
• Imports: igraph, ellipse, corpcor, RColorBrewer, dplyr, tidyr, reshape2, matrixStats, rARPACK, gridExtra, ggrepel, BiocParallel, rgl, rlang
• Install: BiocManager::install("mixOmics")

When to Use

• Performing unsupervised exploratory analysis and dimensionality reduction using pca and spca.
• Integrating multiple omics datasets measured on the same samples (N-integration) using DIABLO.
• Conducting supervised classification and feature selection for multi-class problems using splsda.

When NOT to Use

• For simple univariate differential expression, use limma or DESeq2 instead because mixOmics is designed for multivariate feature selection and integration.
• For raw mass spectrometry or sequencing read preprocessing, use xcms or edgeR for normalization, as mixOmics requires pre-processed, normalized continuous matrices.

Data Requirements

• Input: Numeric data matrix or data frames with $N$ observations (samples) in rows and $P$ predictors (variables) in columns.
• Pre-filtering: Data should be pre-filtered (e.g., < 10K predictors) to remove near-zero variance predictors and reduce computational time.
• Normalization: Data must be normalized prior to analysis (e.g., log-transformed for transcriptomics, CLR-transformed for compositional microbiome data).

Key Parameters

• ncomp (none): The number of principal components or latent dimensions to extract in pca, spca, or splsda.
• scale (TRUE): Logical indicating whether to scale variables to unit variance before analysis in pca or tune.pca.
• center (TRUE): Logical indicating whether to center variables to zero mean in pca.
• keepX (none): A numeric vector specifying the number of variables to select on each component in sparse methods like spca.
• comp (none): Specifies which components to plot in plotIndiv or plotVar (e.g., c(1, 2)).
• group (none): Factor vector used to color samples by class in plotIndiv or biplot.

Best Practices

• Use tune.pca to evaluate the cumulative proportion of explained variance and choose the optimal ncomp via a screeplot.
• Center and scale the data in pca to ensure variables with large variances do not dominate the components.
• Use selectVar to extract and rank the most important variables contributing to each component.
• Visualize sample clustering with plotIndiv and variable correlations with plotVar (correlation circle plot).

Common Pitfalls

• Misinterpreting correlation circle plots on unscaled data: If data is not scaled, cosine angles do not accurately reflect correlations; fix this by ensuring scale = TRUE is set in the PCA/PLS method.
• Overfitting in supervised models: Using PLS-DA without cross-validation can yield perfect separation by chance; fix this by using cross-validation functions like perf and tune.splsda to objectively select ncomp and keepX.
• Incorrect data orientation: mixOmics expects samples in rows and variables in columns; fix this by transposing the matrix if samples are in columns.

Alternatives

• MOFA2: Differs by using Bayesian Group Factor Analysis for multi-omics integration, handling missing values natively.
• RGCCA: Differs by focusing on regularized generalized canonical correlation analysis, which forms the theoretical basis for some mixOmics methods but offers different tuning.

Citations

• Rohart F, Gautier B, Singh A, Lê Cao KA (2017). "mixOmics: An R package for 'omics feature selection and multiple data integration." PLOS Computational Biology, 13(11): e1005752.
• Singh A, Shannon CP, Gautier B, Rohart F, Vacher M, Tebbutt SJ, Lê Cao KA (2019). "DIABLO: an integrative, multi-omics, multivariate method for multi-group classification." Bioinformatics, 35(17): 3055-3062.

References

• Homepage: http://www.mixOmics.org
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/mixOmics/inst/doc/vignettes.html

---

Run this on BioMate

This skill is the knowledge layer — when, why, and how to use mixomics. To run this analysis on your own data with managed compute, automated QC, and reproducible outputs, use BioMate — free to start.

▶ Open mixomics on BioMate →