A related convolutional neural network for cancer diagnosis using microRNA data classification

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNA (miRNA) expression profiles are crucial biomarkers for cancer classification.
• Convolutional Neural Networks (CNNs) have demonstrated efficacy in various pattern recognition tasks.
• Existing classification methods may face challenges with high-dimensional genomic data and computational efficiency.

Purpose of the Study:

• To develop and evaluate a novel, computationally efficient method for cancer classification using miRNA data.
• To enhance classification accuracy by employing a union of two CNNs optimized by a genetic algorithm.
• To compare the proposed method's performance against established classifiers on a large-scale real-world dataset.

Main Methods:

• A convolutional neural network (CNN)-based model optimized by a genetic algorithm was developed.
• The method utilizes a union of two CNNs to leverage inter-network knowledge exchange.
• The approach was tested on a microRNA dataset comprising genomic information from 8129 patients across 29 cancer types.

Main Results:

• The proposed method achieved 100% classification accuracy in 24 out of 29 cancer types.
• In seven specific cases, the method attained 100% accuracy, surpassing previously reported results.
• Performance analysis demonstrated superior accuracy compared to 22 well-known classifiers and 77 previously reported classifiers.

Conclusions:

• The developed CNN-based method offers a highly accurate and computationally efficient solution for cancer classification from miRNA data.
• The union of CNNs and genetic algorithm optimization effectively enhances classification performance.
• This approach holds significant potential for improving diagnostic accuracy in oncology.