D3K: The Dissimilarity-Density-Dynamic Radius K-means Clustering Algorithm for scRNA-Seq Data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Single-cell sequencing data presents significant clustering challenges due to high dimensionality and noise.
• Traditional clustering algorithms like K-means are often unsuitable for the unique characteristics of single-cell data.
• Noise points and inherent data complexity can lead to suboptimal clustering results.

Purpose of the Study:

• To develop an improved clustering algorithm for single-cell sequencing data.
• To address the limitations of the K-means algorithm in handling high-dimensional, noisy datasets.
• To enhance the accuracy and robustness of single-cell data clustering.

Main Methods:

• Proposes a Dissimilarity-Density-Dynamic Radius-K-means clustering algorithm.
• Incorporates a dynamic radius parameter to adaptively adjust based on data characteristics, mitigating noise.
• Calculates cluster weights using dissimilarity density, average contrast, and dissimilarity to determine high-quality initial cluster centers.

Main Results:

• The proposed algorithm demonstrates superior clustering performance on single-cell data compared to existing methods.
• Achieves higher scores across multiple clustering indices, indicating improved accuracy and efficiency.
• Effectively eliminates the influence of noise points and optimizes clustering outcomes.

Conclusions:

• The Dissimilarity-Density-Dynamic Radius-K-means algorithm offers a robust solution for single-cell data clustering.
• Overcomes the inherent randomness and limitations of the traditional K-means algorithm.
• Provides a more effective approach for analyzing complex single-cell sequencing datasets.