Network enhancement as a general method to denoise weighted biological networks

Bo Wang,Marinka Zitnik,Armin Pourshafeie,Junjie Zhu,Carlos D Bustamante,Serafim Batzoglou,Jure Leskovec

doi:10.1038/s41467-018-05469-x

Abstract

Networks are ubiquitous in biology where they encode connectivity patterns at all scales of organization, from molecular to the biome. However, biological networks are noisy due to the limitations of measurement technology and inherent natural variation, which can hamper discovery of network patterns and dynamics. We propose Network Enhancement (NE), a method for improving the signal-to-noise ratio of undirected, weighted networks. NE uses a doubly stochastic matrix operator that induces sparsity and provides a closed-form solution that increases spectral eigengap of the input network. As a result, NE removes weak edges, enhances real connections, and leads to better downstream performance. Experiments show that NE improves gene–function prediction by denoising tissue-specific interaction networks, alleviates interpretation of noisy Hi-C contact maps from the human genome, and boosts fine-grained identification accuracy of species. Our results indicate that NE is widely applicable for denoising biological networks.

Highlights

Networks are ubiquitous in biology where they encode connectivity patterns at all scales of organization, from molecular to the biome
We evaluate the network denoised by Network Enhancement (NE) against the same network denoised by alternative methods: network deconvolution (ND)[10] and diffusion state distance (DSD)[11]
While ND and DSD are denoising algorithms, MU is a feature learning algorithm that learns low-dimensional representations for nodes based on their steady-state topological positions in the network

Summary

Introduction

Networks are ubiquitous in biology where they encode connectivity patterns at all scales of organization, from molecular to the biome. The main crux of NE is the observation that nodes connected through paths with high-weight edges are more likely to have a direct, high-weight edge between them[12,13] Following this intuition, we define a diffusion process that uses random walks of length three or less and a form of regularized information flow to denoise the input network (Fig. 1a and Methods). We define a diffusion process that uses random walks of length three or less and a form of regularized information flow to denoise the input network (Fig. 1a and Methods) This diffusion generates a network in which nodes with strong similarity/ interactions are connected by high-weight edges while nodes with weak similarity/interactions are connected by low-weight edges (Fig. 1b). NE has an efficient and easy to implement closed-form solution for the diffusion process, and provides mathematical guarantees for this converged solution. (Fig. 1b and Methods)

Methods

Results

Conclusion