Analyzing large-scale genomic data, such as from an RNA-sequencing experiment, often yields long gene lists. To make sense of these lists, researchers use functional enrichment analysis. This method identifies biological themes or processes that are over-represented in a gene set. The R package clusterProfiler, part of the Bioconductor project, automates this process, transforming extensive gene lists into meaningful biological insights.
Core Functional Analysis Types
To uncover biological meaning, clusterProfiler interacts with several annotation databases. Each database offers a different perspective on gene function, allowing for a multi-faceted analysis of the data.
Gene Ontology (GO) Enrichment
Gene Ontology analysis categorizes genes based on a structured vocabulary from the Gene Ontology Consortium. This analysis is divided into three domains: Biological Process (BP), which describes larger biological programs; Molecular Function (MF), which details the elemental activities of a gene product; and Cellular Component (CC), which describes where a gene product is active. GO enrichment reveals which terms are more common in a gene list than expected by chance.
KEGG Pathway Enrichment
The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a database of molecular pathways. KEGG pathway enrichment analysis maps genes onto these diagrams of cellular processes, such as metabolic pathways and signaling cascades. Identifying which pathways are over-represented helps researchers hypothesize about the functional consequences of gene expression changes. clusterProfiler accesses the latest KEGG data for thousands of organisms.
Reactome Pathway Analysis
Reactome is another pathway database providing detailed, reaction-level descriptions of biological processes. A Reactome analysis offers a more granular view of molecular events within a pathway compared to KEGG. Its focus on the step-by-step nature of reactions provides insight into how genes might coordinate to produce a biological effect. The `ReactomePA` package works with clusterProfiler to facilitate this analysis.
Other Databases
The versatility of clusterProfiler extends to other databases. It can perform enrichment analysis using the Disease Ontology (DO), which connects genes to human diseases. It also supports the Molecular Signatures Database (MSigDB), a large collection of annotated gene sets. This flexibility allows users to tailor analysis to specific research questions, like investigating disease mechanisms.
Preparing Gene Lists for Analysis
The primary input for clusterProfiler is a list of gene identifiers. Proper preparation of this list is necessary for reliable results. The package accommodates several gene ID types, including Official Gene Symbols, Ensembl IDs, and Entrez Gene IDs. For many core functions, using Entrez IDs is the most direct route, as many of the underlying annotation databases are organized around this identifier.
A “universe” or background gene set is a component of a statistically robust enrichment analysis. This set should contain all genes measured in the initial experiment, such as all genes detected in an RNA-seq study. The analysis uses this background list to determine if genes of interest are over-represented compared to all genes that could have been selected. This comparison provides the statistical foundation for the calculation.
Researchers often have gene lists with identifiers other than the required Entrez IDs. To address this, clusterProfiler includes the `bitr` helper function. This utility translates gene IDs from one format to another, such as converting Gene Symbols into their corresponding Entrez IDs. This function ensures the gene list is in the correct format before analysis.
Running an Analysis and Interpreting Output
Once the gene list is prepared, analysis involves calling a main function like `enrichGO` for Gene Ontology or `enrichKEGG` for KEGG pathways. These functions require the gene list, organism specification, and the desired ontology or pathway database. The output is a results table containing the enriched terms and associated statistics, which requires careful interpretation.
The `p.adjust` or `qvalue` column is the most important in the output. This value is the p-value corrected for multiple testing. Because an analysis tests thousands of terms simultaneously, this correction is necessary to control the false discovery rate. A lower `p.adjust` value indicates higher confidence that a term’s enrichment is statistically significant.
Two other informative columns are `GeneRatio` and `BgRatio`. The `GeneRatio` displays the proportion of genes from the input list associated with a specific biological term. For example, a 20/100 ratio means 20 of 100 input genes are in that term. The `BgRatio` shows the proportion of genes from the background universe associated with the same term. Comparing these ratios shows the extent of enrichment.
The `Description` column provides a human-readable name for the biological term or pathway, such as “inflammatory response.” This allows for biological interpretation of the results. The table provides a ranked list of biological functions most relevant to the input gene list, guiding the next research steps.
Visualizing Enrichment Results
A feature of clusterProfiler is generating visualizations directly from the analysis results. These plots are instrumental for summarizing and interpreting long lists of enriched biological terms, distilling complex data into intuitive graphical formats.
Simple visualizations include bar plots and dot plots. A bar plot can display the most enriched terms, with bar length corresponding to the number of genes or enrichment significance. Dot plots convey multiple pieces of information; the dot size can represent the number of genes, while the color can represent the adjusted p-value, providing a quick, multi-dimensional overview of the top findings.
For a systems-level view, clusterProfiler offers network-based visualizations. The enrichment map, or `emapplot`, organizes enriched terms as a network where overlapping gene sets create connections between terms. This visualizes relationships and groups redundant terms into functional clusters, making it easier to see broader biological themes.
Another visualization is the concept network, or `cnetplot`. This plot links genes to the biological concepts or pathways they are involved in. It highlights which genes are shared across different enriched terms, revealing the genes that may be driving the biological processes. These plots show the intricate connections within the enrichment results.