LogicGep: Boolean networks inference using symbolic regression from time-series transcriptomic profiling data.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) reconstruction from transcriptomic data is crucial for understanding cellular mechanisms.
• Boolean networks (BNs) offer a qualitative framework for modeling gene interaction dynamics.
• Inferring both network topology and gene interaction logic simultaneously from noisy, discretized gene expression data remains a significant challenge.

Purpose of the Study:

• To develop a novel method, LogicGep, for accurate and efficient inference of Boolean networks from time-series gene expression data.
• To address the challenges of noise and information loss in gene expression data during network inference.
• To improve the simultaneous reconstruction of network topology and gene interaction dynamics.

Main Methods:

• LogicGep formulates Boolean function identification as a symbolic regression problem.
• It employs multi-objective optimization using an improved gene expression programming algorithm.
• A feed-forward neural network is used to select the optimal Boolean function from evolved candidates, balancing dynamic and topological characteristics.

Main Results:

• LogicGep successfully infers accurate Boolean network models from both synthetic and real-world gene expression datasets.
• The method outperforms existing BN inference algorithms in both network topology reconstruction and Boolean function identification.
• LogicGep demonstrates significantly improved computational efficiency, being hundreds of times faster than other methods for large-scale network inference.

Conclusions:

• LogicGep provides a robust and efficient solution for Boolean network inference from gene expression data.
• The method enhances the accuracy of GRN topology and logic reconstruction.
• LogicGep's speed makes it particularly suitable for analyzing large and complex biological networks.