Interactive Naive Bayesian network: A new approach of constructing gene-gene interaction network for cancer...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Naive Bayesian (NB) network classifiers are widely used for DNA microarray data analysis.
• The NB network's strong conditional independence assumption can limit its classification performance.

Purpose of the Study:

• To propose an interactive Naive Bayesian (INB) network to enhance cancer classification accuracy from DNA microarray data.
• To address the limitations of the NB network's conditional independence assumption.

Main Methods:

• Selected differently expressed genes (DEGs) to reduce data dimensionality.
• Identified an interactive parent gene with the most influence for each DEG.
• Calculated weights to represent gene-gene interactions and constructed a gene-gene interaction network.

Main Results:

• The INB network achieved higher classification accuracies compared to the standard NB network.
• Experimental validation was performed on leukemia and colon DNA microarray datasets.
• The INB network visually represents gene-gene interactions.

Conclusions:

• The proposed INB network effectively improves cancer classification accuracy using DNA microarray data.
• INB networks offer a more nuanced approach by incorporating gene-gene interactions.
• This method provides a visually interpretable gene-gene interaction network.