A spectral clustering with self-weighted multiple kernel learning method for single-cell RNA-seq data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Single-cell RNA-sequencing (scRNA-seq) data analysis is crucial in bioinformatics.
• Existing spectral clustering methods for scRNA-seq data often involve multiple steps, potentially leading to information loss and performance degradation.
• Kernel selection significantly impacts the performance of kernel-based methods.

Purpose of the Study:

• To develop a novel clustering method for scRNA-seq data that directly discovers groupings.
• To address limitations of traditional spectral clustering by integrating multiple kernel learning.
• To automatically learn optimal similarity information from scRNA-seq data.

Main Methods:

• Proposed a multiple kernel learning model for scRNA-seq data clustering.
• Developed a method to automatically learn similarity information, transforming candidate solutions towards discrete approximations.
• Utilized standard Support Vector Machine (SVM) solvers for efficient model solution.

Main Results:

• The proposed method effectively discovers groupings in scRNA-seq data.
• Experimental results on benchmark scRNA-seq datasets demonstrate superior performance compared to existing methods.
• The model leverages automatically learned similarity information for improved clustering accuracy.

Conclusions:

• The novel multiple kernel learning approach offers a more effective strategy for scRNA-seq data clustering.
• Automatically learned similarity information enhances clustering performance by better approximating discrete solutions.
• The method provides a robust and efficient tool for scRNA-seq data analysis, with open-source implementation available.