Abstract

   The past decades have brought a steady growth of pathway databases and
   enrichment methods. However, the advent of pathway data has not been
   accompanied by an improvement in interoperability across databases,
   hampering the use of pathway knowledge from multiple databases for
   enrichment analysis. While integrative databases have attempted to
   address this issue, they often do not account for redundant information
   across resources. Furthermore, the majority of studies that employ
   pathway enrichment analysis still rely upon a single database or
   enrichment method, though the use of another could yield differing
   results. These shortcomings call for approaches that investigate the
   differences and agreements across databases and methods as their
   selection in the design of a pathway analysis can be a crucial step in
   ensuring the results of such an analysis are meaningful. Here we
   present DecoPath, a web application to assist in the interpretation of
   the results of pathway enrichment analysis. DecoPath provides an
   ecosystem to run enrichment analysis or directly upload results and
   facilitate the interpretation of results with custom visualizations
   that highlight the consensus and/or discrepancies at the pathway- and
   gene-levels. DecoPath is available at
   [34]https://decopath.scai.fraunhofer.de, and its source code and
   documentation can be found on GitHub at
   [35]https://github.com/DecoPath/DecoPath.

INTRODUCTION

   In recent years, high-throughput (HT) technologies have given rise to a
   perpetual influx of -omics data, requiring pragmatic approaches to sift
   out meaning. One of the most common applications of HT technologies is
   gene expression profiling to simultaneously determine the expression
   patterns of thousands of genes at the transcription level under certain
   conditions ([36]1). While a host of statistical techniques are
   available to identify genes that differ in expression depending on a
   particular condition, gene set or pathway enrichment analysis methods
   represent a major class of tools researchers employ to group lists of
   genes into defined pathways and understand the functional roles of
   genes for any given set of conditions ([37]2). To date, almost a
   hundred different pathway enrichment methods have been proposed,
   including the popular over-representation analysis (ORA) and gene set
   enrichment analysis (GSEA) ([38]3). Though these methods may vary based
   on the overarching categories they fall into (e.g. topology versus
   non-topology-based) or the statistical techniques used, they have
   widely shown their ability to deconvolute biological pathways
   dysregulated in a given state ([39]4).

   Numerous pathway databases have been developed which aim at
   representing biological pathways from various vantage points (e.g.
   differing scopes, contexts, boundaries or pathway types). The existence
   of several hundreds of these databases reflects the inherent complexity
   and variability of biological processes that occur in living organisms
   ([40]5). Further compounding this complexity is the fact that
   biological pathways housed in these databases are human constructs,
   delimited based on abstract boundaries defined by a researcher or the
   consensus of the community. This implies that a well-studied pathway
   could contain different biological entities depending on the boundaries
   defined by the databases that store it. These differences across
   databases can manifest in variability in the results of pathway
   enrichment analysis ([41]6,[42]7), in a similar way as methods can
   impact results ([43]4,[44]8–10).

   Recent approaches to pathway enrichment analysis have focused on the
   integration of multiple datasets across different platforms to ensure a
   broader coverage of significantly enriched pathways ([45]11–13). Other
   techniques attempt to account for potential differences that may arise
   in the results of pathway enrichment analysis by combining gene sets
   from several pathway databases. For instance, ([46]14) presented an
   approach that leverages GSEA to calculate a combined enrichment score
   for multiple -omics layers using several databases. However, performing
   pathway enrichment analysis using multiple databases to increase the
   number of pathways covered can only partially address the challenges
   associated with variability in results. This is because such an
   approach falls short of leveraging the substantial overlap of pathway
   knowledge across databases which could provide more comprehensive
   results ([47]15–17) or shed light on inconsistencies across pathway
   databases ([48]18). Furthermore, combining several databases can result
   in redundant pathways, an issue tackled by the SetRank algorithm which
   discounts significant gene sets if their significance can be explained
   by their overlap with another gene set ([49]19). Finally, a possible,
   natural solution to better connect and structure redundant information
   across databases lies in leveraging pathway ontologies ([50]20) or
   pathway mappings with database cross-references ([51]17). By connecting