Integration strategies of multi-omics data for machine learning analysis

Abstract

Increased availability of high-throughput technologies has generated an ever-growing number of omics data that seek to portray many different but complementary biological layers including genomics, epigenomics, transcriptomics, proteomics, and metabolomics. New insight from these data have been obtained by machine learning algorithms that have produced diagnostic and classification biomarkers. Most biomarkers obtained to date however only include one omic measurement at a time and thus do not take full advantage of recent multi-omics experiments that now capture the entire complexity of biological systems. Multi-omics data integration strategies are needed to combine the complementary knowledge brought by each omics layer. We have summarized the most recent data integration methods/ frameworks into five different integration strategies: early, mixed, intermediate, late and hierarchical. In this mini-review, we focus on challenges and existing multi-omics integration strategies by paying special attention to machine learning applications.

Keywords: Deep learning; Integration strategy; Machine learning; Multi-omics; Multi-view; Network.

Graphical abstract

Fig. 1

Structure of an interpretable artificial neural network. The input layer is followed by an additional pathway layer, where each node corresponds to a known molecular pathway. If a molecule is known to be involved in a pathway, a connection is made between the two. Hence, important pathways implicated in the outcome are activated with bigger weights during training. Figure inspired from Deng et al. (2020).

Fig. 2

Example of a mixed artificial neural network. Each omics block is first reduced to a latent representation using independent Stacked Sparse Autoencoders (SAE). The new representations learned are integrated in a final shared layer. The common representation is used for downstream analysis such as prediction or clustering. Figure inspired from Xu et al. (2019) .