HML

HML performs hierarchical maximum likelihood clustering of high-dimensional biological data to identify subtypes in multiomics and genome-wide association study (GWAS) datasets.

Key Features:

• Hierarchical Framework: Employs a hierarchical maximum likelihood clustering scheme that partitions data while accommodating overlapping groups.
• Dimensionality Handling: Handles high-dimensional settings where the number of features may exceed the number of samples.
• Initialization-Free Process: Operates without requiring initial parameter settings for cluster assignment.
• Derivative-Free Computation: Avoids computation of first and second derivatives of likelihood functions.
• Distribution and Centroid Utilization: Leverages distributional information and centroid-based representations for clustering.

Scientific Applications:

• Multiomics Data Analysis: Integrates diverse omics datasets to uncover complex biological interactions and group samples into subtypes.
• Genome-Wide Association Studies (GWAS): Clusters genotype and phenotype data to aid identification of genetic-variant-associated disease subtypes.

Methodology:

Implements a hierarchical maximum likelihood framework that forms clusters using distributional and centroid information, operates without initialization, avoids computing first and second derivatives, and is applicable to high-dimensional datasets where sample size is smaller than dimensionality.