Similarity Network Fusion (SNF)

Preface

In multi-omics studies, we often measure the same patients in several different ways. RNA sequencing may separate one set of patients clearly, while DNA methylation may highlight a different relationship. These views are not necessarily contradictory. Each assay describes a different part of the biology.

Similarity Network Fusion (SNF) was developed for this type of problem. Instead of putting all measurements into one large table, SNF first asks a simpler question within each dataset: which patients are similar to each other? It then combines these patient-to-patient relationships across datasets.

We use RNA expression and DNA methylation throughout the chapter. We begin with four patients so that the movement of information between the networks can be followed. We then use a larger simulated dataset before going through the mathematics in detail.

To follow the chapter, you only need to be familiar with samples, variables, distances and clustering.

Start with the Introduction.