DeepAutoGlioma: a deep learning autoencoder-based multi-omics data integration and classification tools for glioma...

Area of Science:

• Computational biology
• Genomics
• Machine learning in oncology

Background:

• Glioma subtype classification is crucial for targeted therapy.
• Tumor heterogeneity necessitates integrating multi-omics data for accurate classification.
• Deep learning offers a powerful approach for analyzing complex genomic datasets.

Purpose of the Study:

• To develop a deep learning framework for glioma subtype classification.
• To integrate transcriptome and methylome data for improved diagnostic accuracy.
• To support clinical diagnosis of glioma subtypes.

Main Methods:

• Preprocessing of transcriptome and methylome data.
• Identification of differentially expressed genes and CpGs associated with survival using Cox regression.
• Feature selection and integration using autoencoders for dimensionality reduction.
• Classification of glioma subtypes using Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN).

Main Results:

• CNN outperformed ANN in classifying lower-grade gliomas (LGG) and glioblastoma multiforme (GBM), achieving high accuracy (98.03% for LGG, 94.07% for GBM).
• The developed model demonstrated superior performance compared to random gene-CpG pairs, preprocessed data, and single omics data.
• The integrated multi-omics approach provided robust classification with high precision and sensitivity.

Conclusions:

• A novel feature selection and data integration strategy led to the development of DeepAutoGlioma.
• DeepAutoGlioma is an effective framework for diagnosing glioma subtypes.
• The study highlights the potential of deep learning in precision oncology for glioma classification.