Comparative Analysis of Clustering Algorithms in High-Dimensional Biomedical Applications

High-dimensional data in biomedicine such as single-cell RNA-seq (scRNA-seq), imaging, multi-omics data, pose unique challenges for clustering analysis. In this work, we performed an extensive empirical comparison for a representative set of clustering algorithms, including k-Means, Hierarchical clustering, Gaussian Mixture Models, DBSCAN, spectral clustering, and deep clustering approaches on biomedical data. We tested each algorithm on internal and external metrics as well as stability and computational performance. Our finding suggest that Spectral Clustering, Deep Clustering methods consistently find clusters which are more biologically interpretable and less sensitive, however, all at the cost of higher computational complexity. These results contribute useful insights on when to use which algorithm in highdimensional biomedical applications, trading clustering quality for scalability and interpretability.