Enlightening discriminative network functional modules behind Principal Component Analysis separation in differential-omic science studies

Sara Ciucci,Susanne Sales,Carlo Vittorio Cannistraci,Steven W Poser,Mathias J Gerl,Víctor Jiménez-Jiménez,Jochen Guck,Maik Herbig,Yan Ge,Andrej Shevchenko,Claudio Durán,Andreas Androutsellis-Theotokis,Alessandra Palladini,Luisa María Martínez-Sánchez,Oliver Otto,Yuting Wang

doi:10.1038/srep43946

Abstract

Omic science is rapidly growing and one of the most employed techniques to explore differential patterns in omic datasets is principal component analysis (PCA). However, a method to enlighten the network of omic features that mostly contribute to the sample separation obtained by PCA is missing. An alternative is to build correlation networks between univariately-selected significant omic features, but this neglects the multivariate unsupervised feature compression responsible for the PCA sample segregation. Biologists and medical researchers often prefer effective methods that offer an immediate interpretation to complicated algorithms that in principle promise an improvement but in practice are difficult to be applied and interpreted. Here we present PC-corr: a simple algorithm that associates to any PCA segregation a discriminative network of features. Such network can be inspected in search of functional modules useful in the definition of combinatorial and multiscale biomarkers from multifaceted omic data in systems and precision biomedicine. We offer proofs of PC-corr efficacy on lipidomic, metagenomic, developmental genomic, population genetic, cancer promoteromic and cancer stem-cell mechanomic data. Finally, PC-corr is a general functional network inference approach that can be easily adopted for big data exploration in computer science and analysis of complex systems in physics.

Highlights

Omic sciences are contributing to revolutionize current biomedicine towards a precision and patient-tailored approach
Correlation networks are a very general representation paradigm, which is a point of strength for computational systems biology applications, they can be extracted from any type of omic data regardless of the hypothesis of variable dependency, yet offering a very important information on the internal functional association and organization of the system parts
For a general omic dataset, we developed a data-driven method to construct discriminative correlation networks relying on a prior unsupervised analysis and sample projection in a visualization space by Principal Component Analysis (PCA)

Summary

Introduction

Omic sciences are contributing to revolutionize current biomedicine towards a precision and patient-tailored approach. Correlation networks are a very general representation paradigm, which is a point of strength for computational systems biology applications, they can be extracted from any type of omic data regardless of the hypothesis of variable dependency, yet offering a very important information on the internal functional association and organization of the system parts. Their clear point of weakness is that they cannot describe the internal causality (conditioning or dependency) between the parts of the system. Here, for the first time, we want to propose a method - which we named PC-corr – that uses the PCA loadings to perform unsupervised inference of a linear multivariate-discriminative correlation network

Objectives

Methods

Results

Conclusion