P-CSN: single-cell RNA sequencing data analysis by partial cell-specific network.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell analysis often relies on gene expression, which can be unstable.
• Gene-gene associations offer a more stable alternative for improving downstream analysis.
• Existing conditional cell-specific network (c-CSN) methods struggle with false negatives in networks with strong connections.

Purpose of the Study:

• To develop a novel method for constructing accurate gene-gene association networks in single-cell data.
• To address the false negative problem in network construction inherent in c-CSN methods.
• To introduce a new metric for quantifying cell states and reconstructing cell pseudo-time.

Main Methods:

• Proposed a partial cell-specific network (p-CSN) method utilizing partial independence statistics.
• Developed single-cell network entropy (scNEntropy) based on the p-CSN framework.
• Validated the method's performance on multiple biological datasets.

Main Results:

• The p-CSN method constructs accurate partial cell-specific networks (one cell to one network).
• p-CSN significantly reduces false negatives compared to c-CSN in networks with strong connections.
• Improved gene-gene associations enhance downstream analysis, cell state quantification, and pseudo-time reconstruction.

Conclusions:

• The p-CSN method offers a robust approach for gene-gene association inference in single-cell genomics.
• scNEntropy provides an effective metric for cell state characterization and developmental trajectory inference.
• This work advances computational methods for single-cell data analysis and interpretation.