jSRC

jSRC performs joint dimension reduction and clustering of single-cell RNA-sequencing (scRNA-seq) data to identify cell types and informative genes using sparse representation and optimization.

Key Features:

• Joint Learning Framework: jSRC combines dimension reduction with clustering in a single learning framework to improve scalability and the accuracy of cell-type identification and informative-gene selection.
• Sparse Representation: The algorithm assumes cells of the same type share similar expression patterns, enabling a sparse representation that highlights the most informative genes and enhances interpretability.
• Optimization-Based Approach: jSRC formulates clustering as an optimization problem and derives specific update rules to optimize its objective function.
• Performance and Robustness: Validated on 15 diverse scRNA-seq datasets, jSRC outperforms 12 state-of-the-art methods with an average improvement of 20.29% in evaluated metrics while reducing computational time and demonstrating robustness across tissues and organisms.
• Application to Dynamic Biological Processes: jSRC has been applied to identify dynamic cell types associated with disease progression, including COVID-19.

Scientific Applications:

• Cell Type Identification: Accurate clustering of cells based on gene expression profiles to identify distinct cell types within heterogeneous populations.
• Gene Discovery: Prioritization of informative genes to uncover genetic markers that define cell states or responses.
• Disease Progression Studies: Tracking dynamic changes in cell populations to study disease progression and treatment responses, as demonstrated with COVID-19 data.

Methodology:

jSRC transforms scRNA-seq clustering into an optimization problem, integrates dimension reduction techniques, employs sparse representation to focus on relevant features, and derives update rules to optimize its objective function.