CAEM-GBDT: a cancer subtype identifying method using multi-omics data and convolutional autoencoder network.

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Accurate cancer subtype identification is crucial for effective treatment and prognosis.
• Current methods often rely on single-omics data, limiting comprehensive analysis.
• Integrating multi-omics data offers potential for enhanced cancer subtype classification.

Purpose of the Study:

• To develop an advanced computational method for identifying cancer subtypes using multi-omics data.
• To address the challenge of feature extraction from diverse biological datasets for improved cancer classification.

Main Methods:

• Proposed CAEM-GBDT method integrating gene expression, miRNA expression, and DNA methylation data.
• Utilized a convolutional autoencoder network with a self-attention module for feature extraction.
• Employed Gradient Boosting Decision Tree (GBDT) for the final cancer subtype identification.

Main Results:

• The CAEM-GBDT method demonstrated superior performance compared to existing cancer subtype identification techniques.
• Experimental validation confirmed the effectiveness of the proposed multi-omics approach.
• The study highlights the advantage of combining convolutional autoencoders and GBDT for cancer subtyping.

Conclusions:

• CAEM-GBDT offers a robust and accurate approach for cancer subtype identification.
• Multi-omics data integration significantly enhances classification accuracy.
• The developed method provides a valuable tool for precision oncology research.