adSplit

adSplit performs annotation-driven clustering of microarray gene expression data to identify biologically meaningful patient clusters by integrating functional annotations with expression profiles.

Key Features:

• Annotation-Driven Clustering: Integrates functional annotations to generate candidate gene sets representing specific biological functions for clustering.
• Dynamic Distance Measures: Computes and adjusts distance measures between patient samples based on inclusion or exclusion of annotated gene sets rather than relying solely on static metrics like Euclidean distance.
• Biological Significance Assessment: Applies a resampling-based significance measure to filter and validate resulting clusterings and their associated gene sets.
• Clinical Relevance: Recovers clinically relevant patient subgroups and co-regulated genes from gene expression profiles, enabling stratification by molecular characteristics.
• Unsupervised Discovery Potential: Performs unsupervised clustering using biologically focused gene sets to discover novel disease entities and insights into disease mechanisms.

Scientific Applications:

• Clinical Studies: Identifying co-regulated genes and novel disease subtypes to support patient stratification and personalized medicine approaches.
• Research Exploration: Exploring microarray gene expression data to uncover pathways, interactions, or other biological insights using annotation-driven clustering.

Methodology:

Integrates functional annotations to generate candidate gene sets; computes distance measures between patient samples for each annotated term based on selected gene sets; performs clustering using these dynamic distances and applies a resampling-based significance measure to filter significant clusterings; reports significant clusters with their underlying gene sets and functional definitions.