Dissecting and improving gene regulatory network inference using single-cell transcriptome data.

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Single-cell transcriptome data is crucial for reconstructing gene regulatory networks (GRNs).
• Current GRN inference algorithms struggle to achieve high accuracy, often performing only slightly better than random chance.
• Understanding the limitations of current methods is essential for improving GRN inference.

Purpose of the Study:

• To systematically identify factors limiting the accuracy of GRN inference from single-cell data.
• To evaluate the utility of pre-messenger RNA (pre-mRNA) levels for enhancing GRN inference accuracy compared to mature mRNA levels.
• To validate findings using both simulated and experimental single-cell RNA sequencing (scRNA-seq) datasets.

Main Methods:

• Kinetic modeling and simulation of single-cell data to assess the impact of gene-level and network-level factors on mature mRNA reporting of regulatory activity.
• Utilizing intronic reads from public scRNA-seq datasets as a proxy for pre-mRNA levels.
• Analyzing experimental scRNA-seq data to validate simulation findings and explore factors like transcription factor activity dynamics.

Main Results:

• Mature mRNA levels are often poor indicators of upstream regulatory activities due to inherent biological factors.
• Pre-mRNA levels, proxied by intronic reads, provide a more accurate reflection of regulatory activity compared to mature mRNA levels (exonic reads).
• Inference accuracy using pre-mRNA levels was consistently higher than using mature mRNA levels across simulated and real scRNA-seq data.

Conclusions:

• The study delineates fundamental limitations in current GRN inference methods.
• Incorporating pre-mRNA information significantly enhances the accuracy of GRN inference from single-cell data.
• This work provides a pathway to more reliable GRN reconstruction using scRNA-seq data.