hypercluster

hypercluster automates unsupervised clustering optimization by exploring clustering algorithms and hyperparameters to identify robust groupings in large biological datasets, including single-cell RNA sequencing (scRNA-seq), and is implemented as a Python package and SnakeMake pipeline.

Key Features:

• Automated clustering evaluation: Evaluates multiple clustering models and hyperparameter combinations to generate a range of clustering results.
• Hyperparameter exploration: Automates systematic exploration of hyperparameters that influence clustering outcomes, such as expected number of groups and minimum cluster size.
• Parallelization: Leverages parallel computing to accelerate evaluation across algorithms and parameter sets for large datasets.
• Reproducibility support: Structures analyses to support reproducible clustering configurations and comparisons.
• Implementation: Provided as a Python package with a SnakeMake pipeline for workflow-driven execution.

Scientific Applications:

• Single-cell RNA sequencing (scRNA-seq) clustering: Optimizes clustering parameters to identify and compare cell populations within scRNA-seq datasets.
• Patient-sample stratification: Groups similar patient samples in large clinical datasets using unsupervised clustering.
• High-throughput biological data analysis: Applies automated clustering optimization to large-scale datasets generated by high-throughput techniques.

Methodology:

Automated exploration and evaluation of multiple clustering algorithms and their hyperparameters with parallelized execution, implemented within a Python package and SnakeMake pipeline.