MondrianMap: Navigating Gene Set Hierarchies with Multi-Resolution Enrichment Maps.

Area of Science:

• Bioinformatics
• Systems Biology
• Genomics

Background:

• Gene enrichment analysis is crucial for interpreting large-scale gene expression data.
• Standard enrichment outputs are flat ranked lists, limiting the ability to view biological processes at different abstraction levels or compare directional changes between conditions.
• Contemporary systems biology requires tools that preserve hierarchical structure and encode directionality for complex comparisons.

Purpose of the Study:

• To introduce MondrianMap, an interactive web application for multi-resolution visualization of gene enrichment results.
• To enable the recognition of complex biological patterns, such as directional oppositions and temporal dynamics, that are obscured in traditional flat outputs.
• To facilitate hierarchical visual reasoning and complement statistical enrichment reporting for biological discovery.

Main Methods:

• MondrianMap organizes Gene Ontology (GO) terms into 13 semantically principled layers based on the GOALS framework.
• Enrichment results are rendered as color-encoded rectangular maps where block area signifies significance and color indicates effect direction.
• The application provides an interactive interface for navigating these layers, preserving semantic relations through spatial proximity.

Main Results:

• Case studies using LINCS CRISPR, GTEx aging, and MoTrPat exercise data demonstrate MondrianMap's utility.
• LINCS data revealed opposite enrichment maps for TP53 and KRAS knockouts concerning immune recruitment processes.
• GTEx data visualized the inflammaging paradox, showing distinct regulation of antimicrobial defense in aging blood versus liver.
• MoTrPat data captured temporal dynamics in exercise response, illustrating threshold switches and metabolic shifts in different tissues.

Conclusions:

• MondrianMap offers multi-resolution enrichment visualization for Gene Ontology, enhancing biological interpretation.
• Layer-specific views effectively reveal directional oppositions and multi-scale biological narratives difficult to discern in flat term lists.
• The tool demonstrated utility across cancer, aging, and exercise datasets, supporting hypothesis generation and cross-dataset discovery.