Pathway Enrichment Analysis

A rank-based, non-parametric method that evaluates whether a predefined set of genes shows statistically significant, concordant differences between two biological states. Unlike over-representation analysis, GSEA operates on a ranked list of all genes without requiring an arbitrary significance cutoff.

• Enrichment Score (ES): A running-sum statistic that increases when a gene is in the set and decreases otherwise
• Normalized Enrichment Score (NES): Adjusts the ES for variation in gene set size
• False Discovery Rate (FDR): Controls for multiple hypothesis testing across hundreds of gene sets

The method is particularly sensitive to coordinated expression changes where individual genes may not pass significance thresholds but the pathway as a whole is perturbed.

A statistical approach that uses a hypergeometric test or Fisher's exact test to determine whether a pathway contains more differentially expressed genes than expected by chance. ORA requires a binary classification of genes as 'significant' or 'not significant' based on a threshold.

• Input: A list of significantly differentially expressed genes and a background universe
• Null Hypothesis: The pathway genes are randomly distributed in the differential expression list
• Output: A p-value indicating the probability of observing the overlap by chance

ORA is computationally efficient but loses information by dichotomizing continuous expression data, making it less sensitive to subtle but coordinated pathway-level changes.

A category of methods that rank all genes by a differential expression statistic and then assess whether members of a gene set are clustered at the extremes of the ranked list. FCS bridges the gap between ORA and GSEA by using continuous gene-level statistics without requiring a hard cutoff.

• Kolmogorov-Smirnov statistic: Tests if the distribution of gene-set members differs from the background
• Wilcoxon rank-sum test: Assesses whether gene-set ranks are systematically shifted
• MAXMEAN statistic: Identifies gene sets where a subset of members show strong directional changes

FCS methods are robust to the choice of differential expression threshold and capture both up- and down-regulated pathway activity.

An advanced enrichment approach that incorporates the structural information of a pathway—such as gene-gene interactions, signaling directionality, and node centrality—into the statistical model. PTA recognizes that a differentially expressed hub gene has greater biological impact than a peripheral node.

• Impact Factor Analysis: Combines the magnitude of expression change with the topological importance of each gene
• Signaling Pathway Impact Analysis (SPIA): Models signal propagation through a pathway graph
• NetGSA: Uses Gaussian graphical models to account for gene-gene covariance structures

PTA methods require curated pathway topologies from resources like KEGG or Reactome and are computationally more intensive than gene-list-based approaches.

A critical step in pathway enrichment analysis that adjusts raw p-values to control for the inflation of Type I errors when testing hundreds or thousands of pathways simultaneously. Without correction, the probability of false positives becomes unacceptably high.

• Bonferroni Correction: Divides the significance threshold by the number of tests; highly conservative
• Benjamini-Hochberg FDR: Controls the expected proportion of false discoveries among rejected hypotheses
• Permutation-based FDR: Empirically estimates the null distribution by permuting sample labels

FDR is the standard in enrichment analysis because it balances statistical rigor with discovery power, particularly when pathways share overlapping gene memberships.

The biological interpretation of enrichment results depends entirely on the quality and scope of the underlying gene set annotations. These curated collections define the functional groupings against which experimental data is tested.

• Gene Ontology (GO): Hierarchical annotations for Biological Process, Molecular Function, and Cellular Component
• KEGG: Manually drawn pathway maps with explicit molecular interaction networks
• Reactome: Detailed, peer-reviewed pathway diagrams with reaction-level granularity
• MSigDB: The Molecular Signatures Database, containing over 30,000 gene sets including oncogenic signatures and immunological profiles
• WikiPathways: A community-curated, open-source pathway resource

Selecting the appropriate database requires balancing coverage breadth against annotation depth for the specific biological context.

A statistical method that uses a hypergeometric or Fisher's exact test to determine if a pathway contains more differentially expressed genes than expected by chance. ORA requires a pre-defined threshold to create a binary list of 'significant' and 'non-significant' genes. While computationally simple, it is sensitive to the chosen cutoff and treats all passing genes as equally important, ignoring the magnitude of their change.

A critical post-hoc adjustment applied to p-values when testing hundreds of pathways simultaneously. Without correction, the probability of false positives inflates dramatically. Common methods include:

• Bonferroni correction: Controls the Family-Wise Error Rate (FWER) by dividing alpha by the number of tests.
• Benjamini-Hochberg procedure: Controls the False Discovery Rate (FDR), the expected proportion of false positives among rejected hypotheses, and is the standard for genomic studies.

A structured, species-agnostic bioinformatics resource that provides a controlled vocabulary of gene functions across three independent domains: Molecular Function (the elemental activity of a gene product), Biological Process (a series of molecular events), and Cellular Component (the location where a gene product is active). GO terms form a directed acyclic graph, with child terms being more specific than parent terms, enabling enrichment analysis at multiple levels of functional abstraction.

A core output of GSEA that identifies the subset of genes within a gene set that contributes most to the enrichment signal. The leading-edge subset comprises the genes that appear in the ranked list before the running sum reaches its maximum deviation from zero. This analysis pinpoints the key drivers of a pathway's differential expression, filtering out peripheral members and providing a refined biological hypothesis for follow-up validation experiments.

A comprehensive, community-curated collection of annotated gene sets maintained by the Broad Institute for use with GSEA. MSigDB organizes thousands of gene sets into major collections, including Hallmark gene sets (coherent, well-characterized biological states), C2 curated gene sets (from online pathway databases and literature), and C5 ontology gene sets (derived from GO terms). Using a standardized, vetted database ensures reproducibility across studies.