CONSTRUCTING GENE REGULATORY NETWORK USING CHATTERJEE'S RANK CORRELATION WITH SINGLE-CELL TRANSCRIPTOMIC DATA.

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Understanding gene regulatory networks (GRNs) is crucial for cellular function.
• Existing methods for GRN inference from single-cell RNA sequencing (scRNA-seq) data face limitations due to strong assumptions or high computational costs.
• There is a need for transparent, scalable, and efficient methods for GRN discovery.

Purpose of the Study:

• To introduce a novel multiple testing framework for inferring GRNs from scRNA-seq data.
• To overcome the limitations of existing GRN inference methods.
• To provide a computationally efficient and scalable alternative to complex machine learning models.

Main Methods:

• Utilized Chatterjee's rank correlation coefficient, a nonparametric measure of dependence, for GRN inference.
• Developed a data-driven algorithm to estimate robust testing cutoffs, addressing non-independent observations in scRNA-seq.
• Proposed a new test for directed regulation by exploiting the asymmetric nature of Chatterjee's correlation.

Main Results:

• The proposed method demonstrates superior performance in recovering true regulatory links compared to state-of-the-art approaches.
• Successfully applied to both simulated and real scRNA-seq datasets.
• Enabled the construction of biologically meaningful and directionally informed GRNs.

Conclusions:

• The new framework offers a transparent, scalable, and computationally efficient approach to GRN inference.
• The method effectively handles challenges specific to scRNA-seq data, such as non-independent observations.
• Provides a powerful tool for dissecting complex GRNs and advancing cellular function understanding.