Two-Tier Mapper, an unbiased topology-based clustering method for enhanced global gene expression analysis

Area of Science:

• Computational Biology
• Bioinformatics
• Data Science

Background:

• Existing clustering methods often require user-defined parameters, leading to bias and instability.
• Current methods struggle with complex, high-dimensional datasets and small sample sizes.
• Topological Data Analysis (TDA) offers a novel approach to uncover hidden structures in data.

Purpose of the Study:

• To develop an unbiased and stable clustering method for analyzing complex gene expression datasets.
• To address the limitations of existing clustering techniques, especially for small or variable biological samples.
• To leverage TDA for enhanced data exploration and subgroup identification.

Main Methods:

• Developed Two-Tier Mapper (TTMap), a novel topology-based clustering algorithm.
• TTMap utilizes a Mapper-based topological approach at global and local tiers.
• The method incorporates data-driven parameter selection to minimize user bias.

Main Results:

• TTMap effectively discerns divergent features, identifies outliers, and adjusts for control group variations.
• The method demonstrates superior sensitivity and stability compared to current clustering approaches on synthetic and biological data.
• TTMap's performance is robust, unaffected by control sample removal, normalization choices, or data subselection, especially with small sample sizes.

Conclusions:

• TTMap provides an unbiased, stable, and sensitive clustering solution for complex biological data.
• The method's applicability extends to personalized medicine due to its handling of variable samples.
• TTMap is available as an R package in Bioconductor for broader accessibility.