VarSelLCM

VarSelLCM performs model-based clustering with integrated variable selection and missing-data handling for mixed continuous, integer, and categorical datasets to identify relevant features and infer cluster structure.

Key Features:

• Model-based clustering: Implements clustering within a model-based framework using mixture models.
• Mixed-data support: Handles continuous, integer, and categorical variables within the same clustering model.
• Integrated variable selection: Selects relevant features directly during the clustering process to identify variables that contribute to cluster separation.
• Cluster number selection: Determines the optimal number of clusters using classical information criteria.
• Missing-data assumption (MCAR): Assumes missing values are missing completely at random (MCAR) and incorporates this assumption into the analysis.
• Mixture-model imputation: Leverages mixture models to incorporate missing values into clustering and perform imputation within the modeling framework.

Scientific Applications:

• Genomics: Clustering of genomic datasets containing mixed variable types to identify biologically relevant groups and features.
• Proteomics: Analysis of proteomic datasets with continuous and categorical measurements to detect protein expression patterns and discriminative variables.
• Other omics research fields: Application to multi-omics and other biomedical datasets where mixed-data types and missing values are prevalent.

Methodology:

Model-based approaches using mixture models for continuous, integer, and categorical variables; integrated variable selection within the mixture model; selection of the number of clusters via classical information criteria; missing data assumed MCAR and handled within the mixture-model framework for imputation.