Tensorial blind source separation for improved analysis of multi-omic data

Dirk S. Paul,Andrew E. Teschendorff,Klaus Nordhausen,Joni Virta,Han Jing

doi:10.1186/s13059-018-1455-8

Abstract

There is an increased need for integrative analyses of multi-omic data. We present and benchmark a novel tensorial independent component analysis (tICA) algorithm against current state-of-the-art methods. We find that tICA outperforms competing methods in identifying biological sources of data variation at a reduced computational cost. On epigenetic data, tICA can identify methylation quantitative trait loci at high sensitivity. In the cancer context, tICA identifies gene modules whose expression variation across tumours is driven by copy-number or DNA methylation changes, but whose deregulation relative to normal tissue is independent of such alterations, a result we validate by direct analysis of individual data types.

Highlights

Omic data is most often generated in a multidimensional context
We consider two modified versions of these, whereby tensorial principal component analysis (PCA) is applied as a noise reduction step prior to implementing tensorial independent component analysis (tICA), resulting in two algorithms we call tWFOBI and tWJADE (‘Methods’)
We tested the two tICA algorithms, as well as tensorial PCA, on simulated multi-way data consisting of two different data matrices defined over the same 1000 features and 100 samples (‘Methods’)

Summary

Introduction

Omic data is most often generated in a multidimensional context. For the same individual and tissue type, one may measure different data modalities (e.g. genotype, mutations, DNA methylation or gene expression), which may help pinpoint disease-driver genes [1]. For the same individual, the same data type may be measured across different tissues or cell types [2, 3], which may help identify the most relevant cell types or tissues for understanding disease aetiology. We refer to all of these types of multi-dimensional data generally as multi-way or multi-omic data, and when samples and molecular features are matched, the data can be brought into the form of a multi-dimensional array, formally known as a tensor [4].

Methods

Results

Conclusion