HAL-x

HAL-x performs hierarchical density clustering of high-dimensional biological data to identify multi-resolution cell populations and produce multiple cluster label sets for single-cell and other biomedical analyses.

Key Features:

• Hierarchical Density Clustering: Implements a hierarchical density clustering algorithm that constructs cluster hierarchies directly from raw single‑cell data using supervised linkage methods.
• Supervised Linkage Methods: Uses supervised linkage to guide the formation of hierarchical clusters from density-based structures.
• Multiple Cluster Sets Generation: Generates multiple sets of clusters to assess varying levels of cluster specificity and resolve subpopulations such as leukocyte subtypes including T cells.
• Computational Efficiency: Enables rapid prediction of multiple label sets and scalable analysis of large datasets.
• Tunable Parameters: Provides tunable parameters to adjust clustering specificity for different research needs.

Scientific Applications:

• Single-Cell Data Analysis: Groups individual cells into populations based on cellular profiles to study clinical status, disease progression stages, and drug responses.
• Immune Cell Subtype Resolution: Distinguishes complex immune cell subtypes such as leukocytes and T cells within heterogeneous populations.
• Clinical Status Classification: Applies cluster-derived labels for classification tasks and has demonstrated near‑perfect F1-scores in classifying clinical statuses from single-cell profiles.

Methodology:

HAL-x employs a hierarchical density clustering approach with supervised linkage methods to build cluster hierarchies directly from raw single-cell data, supports tunable parameters, and enables rapid prediction of multiple sets of labels for scalable analysis.