flowPeaks

flowPeaks performs unsupervised clustering of flow cytometry data to identify cellular subpopulations in high-dimensional datasets by locating local peaks in a smoothed density function.

Key Features:

• High-Dimensional Data Compatibility: Handles high-dimensional flow cytometry data where histogram-based approaches are computationally infeasible.
• Identification of Irregular Shape Clusters: Combines finite mixture model concepts and spatial exploration techniques to detect clusters with irregular shapes.
• Automatic and Fast Processing: Begins with an initial K-means partitioning using a large K to enable automated and rapid downstream processing and density smoothing.
• Robustness to Outliers: Maintains cluster integrity and reliable results in the presence of outliers.
• Exhaustive Peak Search: Performs an exhaustive search for local peaks in a smoothed density function and clusters cells based on association to those peaks.

Scientific Applications:

• Subpopulation identification: Identification and classification of cellular subpopulations within flow cytometry datasets.
• Benchmarking and comparison: Performance evaluation and comparison with algorithms such as Misty Mountain, FLOCK, flowMeans, flowMerge, and FLAME.

Methodology:

Initial partitioning with K-means using a large K, generation and smoothing of a density function via finite mixture model concepts and spatial exploration of histograms, exhaustive search for local peaks in the smoothed density, and clustering by assigning cells to peaks.