BARTMAP: a viable structure for biclustering.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Clustering is vital for analyzing messenger RNA (mRNA) and microRNA expression profiles in high-throughput studies.
• Traditional clustering faces limitations with uncorrelated genes or unrelated samples, hindering accurate analysis.
• Biclustering addresses these issues by simultaneously clustering genes and samples, integrating feature selection.

Purpose of the Study:

• To propose and demonstrate an efficient biclustering algorithm using a neural-based classifier.
• To address the computational complexity challenges in identifying local relations in gene expression data.
• To enhance the quality and efficiency of clustering for cancer diagnosis and gene function identification.

Main Methods:

• Modification of the ARTMAP neural-based classifier to perform biclustering.
• Development of the Biclustering ARTMAP (BARTMAP) algorithm.
• Experimental validation on multiple human cancer datasets.

Main Results:

• BARTMAP achieves higher quality clustering structures compared to common biclustering and clustering algorithms.
• The algorithm demonstrates fast run times, making it computationally efficient.
• Successful application on human cancer datasets highlights its practical utility.

Conclusions:

• BARTMAP offers an effective and efficient solution for biclustering gene expression data.
• The algorithm shows significant potential for advancing cancer research, gene function identification, and regulatory network inference.
• Neural-based approaches can overcome computational challenges in complex biological data analysis.