Neuroevolution as a tool for microarray gene expression pattern identification in cancer research.

Area of Science:

• Bioinformatics
• Machine Learning
• Cancer Biology

Background:

• Microarray analysis is crucial for cancer research but faces challenges in identifying expression patterns.
• Extracting meaningful insights from high-dimensional microarray data remains a significant hurdle.

Purpose of the Study:

• To develop a novel approach using Neuroevolution for simultaneous classification of microarray data and selection of relevant genes.
• To adapt the FS-NEAT algorithm with structural operators suitable for high-dimensional genomic data.

Main Methods:

• Utilized a rigorous filtering and preprocessing protocol for selecting 13 quality microarray datasets across three cancer types.
• Adapted the FS-NEAT algorithm, incorporating new structural operators for high-dimensional data analysis.
• Employed Neuroevolution, combining neural networks and evolutionary computation, for gene selection and sample classification.

Main Results:

• Neuroevolution successfully classified microarray samples, outperforming existing methods.
• Identified 177 genes, with 82 validated for their respective cancer types and 44 for other cancers, suggesting potential biomarkers.
• Detected five long non-coding RNAs, four with unknown functions, and revealed expression patterns linked to extracellular matrix, exosomes, and cell proliferation.

Conclusions:

• Neuroevolution offers a powerful tool for analyzing cancer microarray data, enhancing gene identification and classification.
• The identified genes and long non-coding RNAs present novel targets for cancer biomarker discovery and therapeutic exploration.
• This approach contributes to unraveling complex molecular mechanisms in tumorigenesis.