scHOIS: Determining Cell Heterogeneity Through Hierarchical Clustering Based on Optimal Imputation Strategy

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-throughput cellular heterogeneity analysis.
• scRNA-seq data are characterized by sparsity and noise due to dropout events, complicating downstream analyses.
• Robust computational methods are needed to address data limitations in scRNA-seq.

Purpose of the Study:

• To develop a novel computational algorithm for single-cell heterogeneity analysis.
• To address data sparsity and improve clustering accuracy in scRNA-seq data.
• To provide a flexible and accurate two-stage imputation and clustering strategy.

Main Methods:

• A two-stage algorithm, scHOIS, was developed for single-cell heterogeneity analysis.
• Stage 1: Masked non-negative matrix factorization for data approximation with optimal rank determination.
• Stage 2: Hierarchical clustering using Pearson correlation on imputed data, with optimal cluster number selection.

Main Results:

• scHOIS effectively and robustly distinguishes cellular differences in scRNA-seq data.
• The algorithm demonstrated superior clustering performance compared to state-of-the-art methods.
• Experiments on real-world datasets validated the efficacy of scHOIS.

Conclusions:

• scHOIS offers an effective solution for analyzing sparse and noisy scRNA-seq data.
• The proposed imputation and hierarchical clustering strategy enhances cellular heterogeneity analysis.
• scHOIS represents a significant advancement in computational tools for scRNA-seq data interpretation.