Single-cell gene regulatory network analysis for mixed cell populations.

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene regulatory networks (GRNs) govern cellular function but are challenging to infer from single-cell RNA sequencing (scRNA-seq) data.
• Existing methods often cluster cells first, ignoring uncertainty and potentially leading to inaccurate GRN estimations.

Purpose of the Study:

• To develop a novel computational method for accurate GRN inference from mixed scRNA-seq data.
• To jointly estimate GRNs for different cell types within a single analysis framework.

Main Methods:

• Developed the variational mixture Poisson log-normal (VMPLN) model for GRN inference.
• Utilized variational inference to handle intractable optimization problems in the mixture Poisson log-normal (MPLN) model.
• Compared VMPLN against state-of-the-art methods using simulations and real scRNA-seq data.

Main Results:

• VMPLN demonstrated superior performance in GRN inference compared to existing methods, particularly with highly mixed cell populations.
• Benchmarking on real data confirmed VMPLN's ability to provide more accurate network estimations.
• Application to SARS-CoV-2 data revealed critical differences in immune cell GRNs between moderate and severe COVID-19 cases.

Conclusions:

• VMPLN offers a robust and accurate approach for inferring gene regulatory networks from mixed scRNA-seq data.
• The method provides valuable insights into cell-type-specific regulatory mechanisms, as demonstrated in the context of COVID-19.
• VMPLN advances the field of single-cell GRN analysis and has implications for understanding disease pathogenesis.