mixOmics: An R package for 'omics feature selection and multiple data integration.
Rohart F, Gautier B, Singh A, Lê Cao KA. · 2017
Read mixOmics: An R package for 'omics feature selection and multiple data integration. by Rohart F, Gautier B, Singh A, Lê Cao KA. (2017) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5687754.txt.
Introduction
Rohart and co-authors present a software paper in PLoS Computational Biology. It introduces mixOmics, an R package for omics data. High-throughput tools now yield many large biological datasets. Transcript, protein, and metabolite data often sit side by side. Most tools test each feature alone, one dataset at a time. That misses links between features and across datasets. mixOmics instead uses multivariate projection methods. It cuts dimensions and picks out molecular signatures.
Two newer frameworks extend the core approach. DIABLO merges several omics types measured on the same samples. MINT merges separate studies that share the same predictors. Both aim at robust signatures for classification and prediction. Worked examples show the package in use.
Catalog record
| CREATOR | Rohart F, Gautier B, Singh A, Lê Cao KA. |
|---|---|
| YEAR | 2017 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
mixOmics: An R package for ‘omics feature selection and multiple data integration By Florian Rohart, Benoît Gautier, Amrit Singh, Kim-Anh Lê Cao, Dina Schneidman Published in: PLoS Computational Biology DOI: 10.1371/journal.pcbi.1005752 ABSTRACT Abstract The advent of high throughput technologies has led to a wealth of publicly available ‘omics data coming from different sources, such as transcriptomics, proteomics, metabolomics. Combining such large-scale biological data sets can lead to the discovery of important biological insights, provided that relevant information can be extracted in a holistic manner. Current statistical approaches have been focusing on identifying small subsets of molecules (a ‘molecular signature’) to explain or predict biological conditions, but mainly for a single type of ‘omics. In addition, commonly used methods are univariate and consider each biological feature independently. We introduce mixOmics, an R package dedicated to the multivariate analysis of biological data sets with a specific focus on data exploration, dimension reduction and visualisation. By adopting a systems biology approach, the toolkit provides a wide range of methods that statistically integrate several data sets at once to probe relationships between heterogeneous ‘omics data sets. Our recent methods extend Projection to Latent Structure (PLS) models for discriminant analysis, for data integration across multiple ‘omics data or across independent studies, and for the identification of molecular signatures. We illustrate our latest mixOmics integrative frameworks for the multivariate analyses of ‘omics data available from the package. "This is a PLOS Computational Biology Software paper." INTRODUCTION The advent of novel ‘omics technologies (e.g. transcriptomics for the study of transcripts, proteomics for proteins, metabolomics for metabolites, etc) has enabled new opportunities for biological and medical research discoveries. Commonly, each feature from each technology (transcripts, proteins, metabolites, etc) is analysed independently through univariate statistical methods including ANOVA, linear models or t-tests. However, such analysis ignores relationships between the different features and may miss crucial biological information. Indeed, biological features act in concert to modulate and influence biological systems and signalling pathways. Multivariate approaches, which model features as a set, can therefore provide a more insightful picture of a biological system, and complement the results obtained from univariate methods. Our package mixOmics proposes multivariate projection-based methodologies for ‘omics data analysis as those provide several attractive properties to the data analyst [1]. Firstly, they are computationally efficient to handle large data sets, where the number of biological features (usually thousands) is much larger than the number of samples (usually less than 50). Secondly, they perform dimension reduction by projecting the data into a smaller subspace while capturing and highlighting the largest sources of variation from the data, resulting in powerful visualisation of the biological system under study. Lastly, their relaxed assumptions about data distribution make them highly flexible to answer topical questions across numerous biology-related fields [2, 3]. mixOmics multivariate methods have been successfully applied to statistically integrate data sets generated from difference biological sources, and to identify biomarkers in ‘omics studies such as metabolomics, brain imaging and microbiome [4–9]. We introduce mixOmics in the context of supervised analysis, where the aims are to classify or discriminate sample groups, to identify the most discriminant subset of biological features, and to predict the class of new samples. We further extended our core method sparse Partial Least Square—Discriminant Analysis (sPLS-DA [10]) that was originally developed for the supervised analysis of one data set. Our two novel frameworks DIABLO and MINT focus on the integration of multiple data sets for different biological questions (Fig 1). DIABLO enables the integration of the same biological N samples measured on different ‘omics platforms (N-integration, [11]), while MINT enables the integration of several independent data sets or studies measured on the same P predictors (P-integration, [12]). To date, very few statistical methods can perform N- and P-integration in a supervised context. For instance, N-integration is often performed by concatenating all the different ‘omics data sets [13], which ignores the heterogeneity between ‘omics platforms and mainly highlights one single type of ‘omics. The other common type of N-integration is to combine the molecular signatures identified from separate analyses of each ‘omics [14], which disregards the relationships between the different ‘omics functional levels. With P-integration, statistical methods are often sequentially combined to accommodate or correct for technical differences (‘batch effects’) among studies before classifying samples with a suitable classification method. Such sequential approaches are time consuming and are prone to overfitting when predicting the class of new samples [12]. Our two frameworks model relationships between different types of ‘omics data (N-integration) or integrate independent ‘omics studies to increase sample size and statistical power (P-integration). Both frameworks aim at identifying biologically relevant and robust molecular signatures to suggest novel biological hypotheses. [Figure: Overview of the mixOmics multivariate methods for single and integrative ‘omics supervised analyses.X denote a predictor ‘omics data set, and y a categorical outcome response (e.g. healthy vs. sick). Integrative analyses include N-integration with DIABLO (the same N samples are measured on different ‘omics platforms), and P-integration with MINT (the same P ‘omics predictors are measured in several independent studies). Sample plots depicted here use the mixOmics functions (from left to right) plotIndiv, plotArrow and plotIndiv in 3D; variable plots use the mixOmics functions network, cim, plotLoadings, plotVar and circosPlot. The graphical output functions are detailed in Supporting Information S1 Text.] The present article first introduces the main functionalities of mixOmics, then presents our multivariate frameworks for the identification of molecular signatures in one and several data sets, and illustrates each framework in a case study available from the package. The data sets supporting the results of this article are available from the mixOmics R package in a processed format. Sweave, R scripts, full tutorials and reports to reproduce the results from the proposed frameworks are available in the supporting information as well as from our website www.mixOmics.org. DESIGN AND IMPLEMENTATION mixOmics is a user-friendly R package dedicated to the exploration, mining, integration and visualisation of large data sets [1]. It provides attractive functionalities such as (i) insightful visualisations with dimension reduction (Fig 1), (ii) identification of molecular signatures and (iii) improved usage with common calls to all visualisation and performance assessment methods (Supporting Information S1 Text). Data input Different types of biological data can be explored and integrated with mixOmics. Prior to the analysis, we assume the data sets have been normalised using appropriate techniques specific for the type of ‘omics technology platform. The methods can handle molecular features measured on a continuous scale (e.g. microarray, mass spectrometry-based proteomics and metabolomics) or sequenced-based count data (RNA-seq, 16S, shotgun metagenomics) that become “continuous” data after pre-processing and normalisation. We denote X a data matrix of size N observations (rows) × P predictors (e.g. expression levels of P genes, in columns). The categorical outcome y (e.g. sick vs healthy) is expressed as a dummy indicator matrix Y, where each column represents one outcome category and each row indicates the class membership of each sample. Thus, Y is of size N observations (rows) × K categories outcome (columns), see example in Suppl S1 Text. While mixOmics methods can handle large data sets (several tens of thousands of predictors), we recommend pre-filtering the data to less than 10K predictors per data set, for example by using Median Absolute Deviation [15] for