Integrating multi-OMICS data through sparse canonical correlation analysis for the prediction of complex traits: a comparison study

Abstract

Motivation Recent developments in technology have enabled researchers to collect multiple OMICS datasets for the same individuals. The conventional approach for understanding the relationships between the collected datasets and the complex trait of interest would be through the analysis of each OMIC dataset separately from the rest, or to test for associations between the OMICS datasets. In this work we show that integrating multiple OMICS datasets together, instead of analysing them separately, improves our understanding of their in-between relationships as well as the predictive accuracy for the tested trait. Several approaches have been proposed for the integration of heterogeneous and high-dimensional (p ≫ n) data, such as OMICS. The sparse variant of Canonical Correlation Analysis (CCA) approach is a promising one that seeks to penalise the canonical variables for producing sparse latent variables while achieving maximal correlation between the datasets. Over the last years, a number of approaches for implementing sparse CCA (sCCA) have been proposed, where they differ on their objective functions, iterative algorithm for obtaining the sparse latent variables and make different assumptions about the original datasets. Results Through a comparative study we have explored the performance of the conventional CCA proposed by Parkhomenko et al. (2009), penalised matrix decomposition CCA proposed by Witten and Tibshirani (2009) and its extension proposed by Suo et al. (2017). The aforementioned methods were modified to allow for different penalty functions. Although sCCA is an unsupervised learning approach for understanding of the in-between relationships, we have twisted the problem as a supervised learning one and investigated how the computed latent variables can be used for predicting complex traits. The approaches were extended to allow for multiple (more than two) datasets where the trait was included as one of the input datasets. Both ways have shown improvement over conventional predictive models that include one or multiple datasets. Availability https://github.com/theorod93/sCCA Supplementary information Supplementary data and material are available at Bioinformatics online.

Machine learning for multi-omics data integration in cancer

“…sCCA (sparse CCA, general-purpose) is a variation of CCA that imposes additional penalties in modeling so that the number of latent variables can be kept low for better interpretation (Rodosthenous et al, 2020).…”

Collective effects of human genomic variation on microbiome function

“…In this case, the number of genes with expression data and the number of polymorphisms will most likely exceed the number of individuals that can feasibly be sampled. Penalized CCA methods have been developed to overcome this issue 29 – 31 . In penalized CCA, also known as sparse CCA or sCCA, sparsity is induced in the features through penalization, for example using an l 1 penalty (lasso) 32 .…”

$\ell_1$-norm constrained multi-block sparse canonical correlation analysis via proximal gradient descent

“…Variants of mCCA have been applied to various applications for dimension reduction and exploration with high dimensional data, including joint blind source separation (Li et al, 2009), multi-omics data integration (Subramanian et al, 2020;Rodosthenous et al, 2020), neuroimaging (Sui et al, 2012) and others. Researchers seek leading mCCA directions that capture coherent variations across blocks in these applications.…”