scBKAP

scBKAP performs clustering of single-cell RNA sequencing (scRNA-seq) data by reconstructing gene expression with an autoencoder, selecting features with M3Drop, reducing dimensionality with PHATE (MPDR), and applying bisecting K-means to identify cell populations.

Key Features:

• Autoencoder Network: Uses an autoencoder to reconstruct scRNA-seq gene expression values and mitigate dropout in single-cell datasets.
• MPDR (M3Drop and PHATE): Combines M3Drop feature selection with PHATE (Potential of Heat-diffusion for Affinity-based Trajectory Embedding) for feature selection and nonlinear dimensionality reduction.
• Bisecting K-means Clustering Algorithm: Applies a bisecting K-means algorithm to reduced representations to delineate distinct cell clusters.
• Dropout and dimensionality handling: Integrates reconstruction and MPDR steps to address high dropout rates and the curse of dimensionality in scRNA-seq data.
• Benchmarking: Demonstrated performance across 21 public scRNA-seq datasets and simulated datasets, outperforming nine state-of-the-art single-cell clustering methods.

Scientific Applications:

• Single-cell transcriptome profiling: Enables genome-wide transcription profiling at single-cell resolution.
• Cell type and state identification: Facilitates identification of cell types and cellular states by clustering transcriptional profiles.
• Developmental biology: Supports analysis of cellular heterogeneity and lineage relationships in developmental studies.
• Cancer research: Supports characterization of tumor heterogeneity and identification of tumor subpopulations.
• Immunology: Supports profiling of immune cell populations and immune-state heterogeneity.

Methodology:

Reconstruct gene expression with an autoencoder, apply M3Drop for feature selection, use PHATE for dimensionality reduction (MPDR), and cluster cells with bisecting K-means.