Equivalent Change Enrichment Analysis: Assessing Equivalent and Inverse Change in Biological Pathways between Diverse Experiments

Abstract

In silico functional genomics have become a driving force in the way we interpret and use gene expression data, enabling researchers to understand which biological pathways are likely to be affected by the treatments or conditions being studied. There are many approaches to functional genomics, but a number of popular methods determine if a set of modified genes has a higher than expected overlap with genes known to function as part of a pathway (functional enrichment testing). Recently, researchers have started to apply such analyses in a new way: to ask if the data they are collecting show similar disruptions to biological functions compared to reference data. Examples include studying whether similar pathways are perturbed in smokers vs. users of e-cigarettes, or whether a new mouse model of schizophrenia is justified, based on its similarity in cytokine expression to a previously published model. However, there is a dearth of robust statistical methods for testing hypotheses related to these questions and most researchers resort to ad hoc approaches. In this work, we propose a statistical approach to answering such questions. First, we propose a statistic for measuring the degree of equivalent change in individual genes across different treatments. Using this statistic, we propose applying gene set enrichment analysis to identify pathways enriched in genes that are affected in similar or opposing ways across treatments. We evaluate this approach in comparison to ad hoc methods on a simulated dataset, as well as two biological datasets and show that it provides robust results.