Boolean networks using the chi-square test for inferring large-scale gene regulatory networks

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Boolean network (BN) modeling is crucial for gene regulatory network (GRN) inference from time-series microarray data.
• Conventional BN methods face significant computational challenges, limiting their scalability for large gene networks.

Purpose of the Study:

• To introduce a novel variable selection method to accelerate BN computation.
• To enhance the efficiency and accuracy of GRN construction using chi-square statistics.

Main Methods:

• A chi-square testing (CST) based variable selection approach was developed for BN modeling.
• The CST-based BN method was evaluated against original BN algorithms using simulated and real yeast cell cycle gene expression data.

Main Results:

• The CST-based BN method significantly reduced computation time compared to original BN algorithms (70.8x and 7.6x faster for specific error levels).
• Network construction accuracy was comparable to full-search BN methods, with a potentially lower false positive rate.
• The method demonstrated efficient inference of large-scale gene regulatory network mechanisms.

Conclusions:

• The CST-based BN method offers a substantial improvement in computational efficiency for BN algorithms.
• This approach enables the effective inference of large-scale gene regulatory networks.