Inferring Gene Regulatory Networks From Single-Cell Transcriptomic Data Using Bidirectional RNN.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate inference of gene regulatory networks (GRNs) is crucial for understanding cellular processes.
• Existing computational methods often decompose GRN inference into subproblems, limiting scalability to large datasets.
• Time-series single-cell RNA-seq (scRNA-seq) data offers rich temporal information for GRN inference.

Purpose of the Study:

• To develop a novel computational algorithm, BiRGRN, for inferring GRNs from time-series scRNA-seq data.
• To address the limitations of existing methods by enabling simultaneous detection of causal relationships.
• To provide a scalable and accurate tool for analyzing complex gene regulatory dynamics.

Main Methods:

• BiRGRN utilizes a bidirectional recurrent neural network (RNN) to model complex, non-linear, and dynamic gene expression relationships.
• The algorithm transforms GRN inference into a regression problem, predicting future gene expression from past data.
• Strategies including bidirectional structure integration and prior knowledge filtering enhance accuracy and stability.

Main Results:

• BiRGRN was validated on four simulated and three real scRNA-seq datasets.
• Comprehensive comparisons demonstrated superior performance against state-of-the-art techniques.
• The method successfully infers GRNs simultaneously from time-series scRNA-seq data.

Conclusions:

• BiRGRN offers an effective and accurate approach for inferring gene regulatory networks from time-series scRNA-seq data.
• The bidirectional RNN architecture captures complex regulatory dynamics.
• The algorithm provides a valuable tool for advancing systems biology and understanding cellular mechanisms.